Mouse Aging Cell Atlas enables discovery of cellular-specific aging changes in different cell types and organs – National Institute on Aging

Aging is a significant risk factor for developing many chronic diseases, and gaining a better understanding of the genetic, molecular, and cellular processes responsible for the aging process is vital to improving quality of life. To this end, two studies, funded in part by NIA and published in Nature, used mouse models to catalog and identify multiple cell-specific and cellular composition changes in different cell types and organs related to the aging process.

The first study describes the creation of a single-cell transcriptomic mouse cell atlas that captures cell-type-specific hallmarks across the lifespan of the mouse. The Tabula Muris Senis or Mouse Aging Cell Atlas provides molecular information on age-related changes in specific cell types across 23 tissues and organs. In the second study, using data from the atlas, researchers found that not only do cell-specific changes occur across multiple cell types and organs, but age-related changes also occur in the cellular composition of different organs.

To create the mouse atlas, researchers performed single-cell RNA sequencing on more than 350,000 cells from male and female mice ranging from one to 30 months old, which models the human aging process from infancy to approximately 100 years old. Data were collected for mice in six age groups at 1, 3, 18, 21, 24 and 30 months. By analyzing multiple organs from the same mouse over that time span, researchers were able to obtain data controlled for age, environment, and epigenetic effects.

Researchers observed that changes in the relative composition of a given cell type with age are more meaningful than comparing proportions of different cell types at a single age. In one analysis of their data, they used the Genome Analysis ToolKit to identify specific gene mutations across all samples simultaneously. They focused on genes that were expressed in at least 75% of cells for each age group within a particular tissue and observed an age-related increase in mutations across all the organs they analyzed. This supports other studies indicating that genomic instability is a hallmark of aging and suggests it occurs in many organs of the body.

In the second study, researchers performed bulk RNA sequencing of proteins and data from the mouse atlas to demonstrate a progression of aging both within and among different organs.

They measured plasma proteins and sequenced RNA from 17 isolated organ types from male and female mice from one month old to maturity (3-6 months old) and aging through adulthood (median 27 months old). Researchers then analyzed whether differentially expressed genes (DEGs) arise and whether they persist with advancing age. Differential gene expression is the activation of different genes within a cell that define its function.

Few DEGs were observed between organs at similar ages but they increased markedly with advancing mouse age, especially when compared with one-month-old mice that were undergoing development. Among their findings on DEGs, they discovered gene expression trajectories that were similar to aging-related processes, including mitochondria dysfunction, impaired protein folding, and inflammation. They also noted that changes in DEGs for common biological pathways in tissues did not seem to be driven by changes in transcription regulatory factors, which turn gene expression on and off. This suggests that additional gene regulatory sites may come into play in the dynamics of DEGs with aging. Further, researchers found that DEGs that began in middle age were highly correlated to similar patterns in later life, indicating that some harmful changes begin earlier in life. The researchers noted that better understanding of these processes may lead to improved interventions to enhance healthspan benefits.

These studies highlight the utility of the Mouse Aging Cell Atlas, as well as the work that can stem from enhanced understanding of aging processes at the cellular, tissue, and organ system level. Future research using these characterizations of aging may help with the development and application of interventions to increase the healthspan and delay aging-related diseases. The mouse atlas data set is available at https://tabula-muris-senis.ds.czbiohub.org/.

This research was supported in part by NIA grants R01-AG045034 and DP1-AG053015.

References: Tabula Muris Consortium. A single-cell transcriptomic atlas characterizes ageing tissues in the mouse. Nature. 2020;583(7817):590-595. doi: 10.1038/s41586-020-2496-1.

Schaum N, et al. Ageing hallmarks exhibit organ-specific temporal signatures. Nature. 2020;583(7817):596-602. doi: 10.1038/s41586-020-2499-y.

Original post:
Mouse Aging Cell Atlas enables discovery of cellular-specific aging changes in different cell types and organs - National Institute on Aging

Building better vaccines for the elderly | The Source – Washington University in St. Louis Newsroom

As human lifespans have gotten longer, certain proteins in our bodies are increasingly prone to take on alternative shapes. These misfolded proteins can ultimately trigger neurodegenerative diseases such as Alzheimers, Parkinsons and Lou Gehrigs disease, formally known as amyotrophic lateral sclerosis (ALS).

Meredith Jackrel, assistant professor of chemistry in Arts & Sciences at Washington University in St. Louis, and her lab group study protein misfolding disorders. They are especially interested in how protein misfolding occurs, how it leads to disease and how scientists might be able to prevent or even reverse protein misfolding. Their work promises applications in flu vaccines as well as in the current coronavirus pandemic.

In this Q&A, Jackrel describes how her labs expertise in protein misfolding and neurodegenerative diseases has made them uniquely qualified to work on developing new amyloid-inspired vaccine technologies aimed at elderly populations.

How does your research relate to the current pandemic?

We are working on the development of new vaccine technologies specifically tailored to elderly populations. We originally initiated this project to evaluate new flu vaccine technologies, but this approach could also be relevant to COVID-19 since seniors are particularly susceptible to its severe complications.

A general problem with vaccination of elderly individuals is immuosenescence, or age-related dysfunction of the immune system. Immunosenescence is typically overcome by the addition of adjuvants to improve immune response and efficacy. However, adjuvants create local inflammation, which obstructs the immune system and makes vaccines less effective.

A colleague at WashU in biomedical engineering, Jai Rudra, studies self-assembling peptides as materials for developing novel vaccines that do not require the use of adjuvants. These self-adjuvanting peptide nanofibers are hypothesized to trigger the autophagy pathway, a kind of cellular recycling that can also promote good immunological functions, which has emerged as a potential vaccine target. Administration of these peptide nanofibers leads to robust, high-affinity, and neutralizing antibody responses without local reactions, making them attractive for vaccine delivery in the elderly.

To further pursue application of these nanofibers, we must now investigate the toxicity and clearance mechanism of these materials.

How are you using your expertise in protein folding/misfolding in your work on vaccine technology?

Peptide nanofiber materials rapidly assemble into configurations that closely resemble the underlying causes of neurodegenerative disorders. These amyloids are recognized as clumps of proteins that accumulate in patients with Alzheimers, Huntingtons and Parkinsons disease.

While there are key differences that we anticipate will not make use of these materials problematic, it is nonetheless essential that the safety and clearance mechanism of the peptide nanofiber vaccines be thoroughly tested. My labs expertise in the development of model systems to study the toxicity and mechanism of disease-associated amyloid proteins is therefore highly relevant to this project.

Furthermore, due to the complexities of studying the peptide nanofibers in mammalian cells, my labs expertise in the use of Bakers yeast as a model system is proving highly relevant for studying the mechanism of clearance of these new materials.

What are your specific goals in this project?

The primary goals for my lab are to determine the toxicity and mechanism of clearance of the peptide nanofiber vaccines in a yeast model system. We aim to compare the toxicity of the nanofibers to the toxicity of disease-associated proteins. We will also employ autophagy-deficient yeast models to establish the mechanism of clearance of the nanofibers. The Rudra lab then aims to assess the efficacy of the nanofiber-based vaccines in aged mice.

Where does the project stand now? What are the next steps?

We have established a yeast model system of these peptide nanofibers and completed much of the preliminary work. Excitingly, we have confirmed that these peptide nanofibers are not toxic in yeast and have made some new insights into their mechanism of clearance. We aim to complete the early stage of this project shortly, and the Rudra lab has begun work in animal systems. Once we complete work with the nanofibers alone, we will begin to test conjugates to various vaccine targets, notably those that underpin COVID-19.

Read this article:
Building better vaccines for the elderly | The Source - Washington University in St. Louis Newsroom

Lattman, Liu, Morrow and Ruhl elected AAAS fellows – UB Now: News and views for UB faculty and staff – University at Buffalo Reporter

Your Colleagues

UBNOW STAFF

Published November 30, 2020

Four UB professors have been elected fellows of the American Association for the Advancement of Science (AAAS), the world's largest general scientific society and publisher of the journal Science.

The honor is bestowed on AAAS members by their peers for their scientifically or socially distinguished efforts to advance science applications. The UB faculty members were among 489 members to receive the prestigious distinction this year.

The new UB fellows include:

AAAS fellows will be recognized in the journal Science on Nov. 27. An induction ceremony will be held during the virtual AAAS Fellows Forum on Feb. 13.

Eaton Lattman (biological sciences)

Lattman was honored for his distinguished contributions in scholarship, education and leadership in the fields of molecular biophysics and structural biology.

A prolific researcher in crystallography and biophysics, Lattman has focused on protein folding and on development and improvement of methods in protein crystallography. He has pioneered the emerging field of using X-ray free electron lasers to study biological and nonbiological processes.

Lattman spent nearly his entire academic career at Johns Hopkins University, as professor of biophysics in both the School of Medicine and the Krieger School of Arts and Sciences, where he also served as dean of research and graduate education. He played a key role in establishing the Hopkins Institute for Biophysical Research.

In 2008, Lattman came to Buffalo to serve as chief executive officer at Hauptman-Woodward Medical Research Institute. He joined the UB Department of Structural Biology in 2009.

In 2013, he was instrumental in the awarding of a $25 million U.S. National Science Foundation grant to UB and its partners to establish BioXFEL, an X-ray laser science center, to transform the field of structural biology. It was UBs first NSF Science and Technology Center Grant. Lattman was named director and led the national consortium until 2017. Under his direction, the consortium made significant progress in refining X-ray laser techniques to study biological processes and innovating new approaches to use these methods to advance materials science and other nonbiological disciplines as well. He continues to serve as a member of the BioXFEL steering committee.

Xiufeng Liu (education)

Liu was recognized for his distinguished contributions to the fields of science education research, and communicating and interpreting science to the public.

Liu is renowned for his scholarship on measuring and evaluating student achievement in science, technology, engineering and math (STEM). He served as the inaugural director of UBs Center for Educational Innovation, with a mission to improve university teaching, learning and assessment.

He also strives to increase scientific literacy among members of the public, and inspired a program at UB called Science and the Public that prepares museum curators, zoo directors, pharmacists and other informal science educators to teach science to a general audience, including by engaging in activities and debates related to science.

Liu has received more than $18 million in research funding, and published more than 100 academic articles and 10 books. He received a doctorate in science education from the University of British Columbia and a masters degree in chemical education from East China Normal University.

Janet Morrow (chemistry)

Morrow was honored for her distinguished contributions to the field of inorganic complexes and their biomedical applications, particularly for magnetic resonance imaging contrast agents and for nucleic acid modifications.

Morrow is an expert in bioinorganic chemistry, with a wide range of innovations and publications in the field. The central theme of her research is the synthesis of inorganic complexes for biomedical diagnostics, sensing or catalytic applications. Focus areas include research and development of novel MRI contrast agents, yeast cell labeling with metal complex probes to track infections, and bimodal imaging agents. Morrow is also an inventor and entrepreneur, having co-founded Ferric Contrast, a startup that is developing iron-containing MRI contrast agents.

She is a recipient of the Jacob F. Schoellkopf Medal presented by the Western New York section of the American Chemical Society, the UB Exceptional Scholar Award for Sustained Achievement, the National Science Foundation Award for Special Creativity and the Alfred P. Sloan Research Fellowship. Morrow holds a doctorate in chemistry from the University of North Carolina at Chapel Hill and a bachelors degree in chemistry from the University of California, Santa Barbara.

Stefan Ruhl (dentistry and oral health sciences)

Ruhl was recognized for his distinguished contributions to the field of oral biology, particularly for work on glycan-mediated microbial adhesion in the oral cavity.

Ruhl is an internationally renowned expert on saliva, oral bacteria and the oral microbiome. His research attempts to unravel the roles that saliva and microorganisms play in health, including in adhesion to the teeth and surfaces of the mouth, defense against pathogens and colonization of the oral cavity. He investigates the molecular mechanisms of microbial binding to glycans, a common but little understood class of biomolecules that help bacteria attach to host surfaces, including those in the mouth. The goal of his lab is to harness tools that ultimately help scientists examine how the microorganisms bind to glycans in the mouth to form dental biofilms more commonly known as plaque increasing the risk for cavities and periodontal disease.

He was among the first researchers to catalogue the human salivary proteome, which is the entirety of proteins present in saliva and in salivary gland ductal secretions. Ruhl has led or participated in recent studies that have identified how saliva is made, tracing each salivary protein back to its source. He also discovered that 2 million years of eating meat and cooked food has led humans to develop a saliva that is now starkly different from that of chimpanzees and gorillas, our closest genetic relatives. This seminal discovery has resulted in collaborative projects exploring saliva to understand the factors that helped shape human evolution and, in particular, the evolution of the human mouth. These evolutionary projects identified a starch-digesting enzyme called amylase in the saliva of dogs and various other starch-consuming mammals, and through analysis of a salivary mucin protein found genetic evidence that humans may have mated with a ghost species of archaic humans.

Ruhl received the 2020 Distinguished Scientist Award in Salivary Research and the 2014 Salivary Researcher of the Year award from the International Association for Dental Research, as well as the UB Exceptional Scholar Award for Sustained Achievement. He holds a doctor of dental surgery degree and a doctoral degree in immunology from Georg-August University of Gttingen.

Here is the original post:
Lattman, Liu, Morrow and Ruhl elected AAAS fellows - UB Now: News and views for UB faculty and staff - University at Buffalo Reporter

New AI That Predicts Shape of Proteins Could Solve 50-Year Problem – The Great Courses Daily News

By Jonny Lupsha, News Writer

According to The New York Times, a lab in London may have developed an artificial intelligence that can do a years-long task in less than a day, solving a longstanding problem of biology. For biologists, identifying the precise shape of a protein often requires months, years, or even decades of experimentation, the article said. It requires skill, intelligence, and more than a little elbow grease.

Now, an artificial intelligence lab in London has built a computer system that can do the job in a few hoursperhaps even a few minutes.

The article said that the laboratory in question, DeepMind, analyzes a string of amino acids that make up a protein and then rapidly and reliably predicts its shape. But why is that so important?

The study of folding proteins began in the 1950s with American biochemist Christian Anfinsen playing a key role.

The first experiments began by taking a protein out of the cell, unfolding it, and then seeing if it could refold in a test tube, independent of any cellular factors, said Dr. Kevin Ahern, Professor of Biochemistry and Biophysics at Oregon State University. The protein Christian Anfinsen picked was the enzyme ribonuclease A, also known as RNase, which turned out to be a serendipitous choice. RNase is relatively small as proteins goabout 100 amino acidsand it is also extraordinarily stable.

Dr. Ahern said that most enzymes are very sensitive to changes in temperature or pH balance, but RNase is not. Anfinsen showed that once an enzyme is unfolded, its capable of refolding outside the cell. His work earned him the 1972 Nobel Prize for Chemistry. Dr. Ahern also said that this process is called renaturation because the protein gets returned to its native or natural state.

Humanity has been studying protein folding for over 60 years. What happens when proteins fold incorrectly? As it turns out, nothing good.

These are the so-called prion diseases, also known as transmissible spongiform encephalopathies or TSEs, Dr. Ahern said. Prion diseases affect humans and other animals. They are a group of degenerative disorders that affect the brain, creating microscopic holes that make the tissue look like a sponge.

He also said that one of the best-known prion diseases is bovine spongiform encephalopathy, also known as Mad Cow Disease. Animals that had it would exhibit behaviors that were consistent with neurological damage, and finding a common cause among them was difficult.

Stanley Prusiner at the University of California at San Francisco ultimately identified the infectious agent as a proteina proteinaceous infectious article he called a prion, Dr. Ahern said. That a protein could be infectious by itself was unheard of at the time. And [it] turned out to be a cellular protein found on the membrane of healthy cells; though its function to this day remains uncertain.

Protein misfolding causes several serious diseases and helps explain why the study of protein folding matters so much.

This article was proofread and copyedited by Angela Shoemaker, Proofreader and Copy Editor for The Great Courses Daily.

Dr. Kevin Ahern contributed to this article. Dr. Ahern is a Professor of Biochemistry and Biophysics at Oregon State University (OSU), where he also received his PhD in Biochemistry and Biophysics.

See more here:
New AI That Predicts Shape of Proteins Could Solve 50-Year Problem - The Great Courses Daily News

Mapping out the mystery of blood stem cells – Science Codex

Princess Margaret scientists have revealed how stem cells are able to generate new blood cells throughout our life by looking at vast, uncharted regions of our genetic material that hold important clues to subtle biological changes in these cells.

The finding, obtained from studying normal blood, can be used to enhance methods for stem cell transplantation, and may also shed light into processes that occur in cancer cells that allow them to survive chemotherapy and relapse into cancer growth many years after treatment.

Using state-of-the art sequencing technology to perform genome-wide profiling of the epigenetic landscape of human stem cells, the research revealed important information about how genes are regulated through the three-dimensional folding of chromatin.

Chromatin is composed of DNA and proteins, the latter which package DNA into compact structures, and is found in the nucleus of cells. Changes in chromatin structure are linked to DNA replication, repair and gene expression (turning genes on or off).

The research by Princess Margaret Cancer Centre Senior Scientists Drs. Mathieu Lupien and John Dick is published in Cell Stem Cell, Wednesday, November 25, 2020.

"We don't have a comprehensive view of what makes a stem cell function in a specific way or what makes it tick," says Dr. Dick, who is also a Professor in the Department of Molecular Genetics, University of Toronto.

"Stem cells are normally dormant but they need to occasionally become activated to keep the blood system going. Understanding this transition into activation is key to be able to harness the power of stem cells for therapy, but also to understand how malignant cells change this balance.

"Stem cells are powerful, potent and rare. But it's a knife's edge as to whether they get activated to replenish new blood cells on demand, or go rogue to divide rapidly and develop mutations, or lie dormant quietly, in a pristine state."

Understanding what turns that knife's edge into these various stem cell states has perplexed scientists for decades. Now, with this research, we have a better understanding of what defines a stem cell and makes it function in a particular way.

"We are exploring uncharted territory," says Dr. Mathieu Lupien, who is also an Associate Professor in the Department of Medical Biophysics, University of Toronto. "We had to look into the origami of the genome of cells to understand why some can self-renew throughout our life while others lose that ability. We had to look beyond what genetics alone can tell us."

In this research, scientists focused on the often overlooked noncoding regions of the genome: vast stretches of DNA that are free of genes (i.e. that do not code for proteins), but nonetheless harbour important regulatory elements that determine if genes are turned on or off.

Hidden amongst this noncoding DNA - which comprise about 98% of the genome - are crucial elements that not only control the activity of thousands of genes, but also play a role in many diseases.

The researchers examined two distinct human hematopoietic stem cells or immature cells that go through several steps in order to develop into different types of blood cells, such as white or red blood cells, or platelets.

They looked at long-term hematopoietic stem cells (HSCs) and short-term HSCs found in the bone marrow of humans. The researchers wanted to map out the cellular machinery involved in the "dormancy" state of long-term cells, with their continuous self-renewing ability, as compared to the more primed, activated and "ready-to-go" short-term cells which can transition quickly into various blood cells.

The researchers found differences in the three-dimensional chromatin structures between the two stem cell types, which is significant since the ways in which chromatin is arranged or folded and looped impacts how genes and other parts of our genome are expressed and regulated.

Using state-of-the-art 3D mapping techniques, the scientists were able to analyze and link the long-term stem cell types with the activity of the chromatin folding protein CTCF and its ability to regulate the expression of 300 genes to control long-term, self-renewal.

"Until now, we have not had a comprehensive view of what makes a stem cell function in a particular way," says Dr. Dick, adding that the 300 genes represent what scientists now think is the "essence" of a long-term stem cell.

He adds that long-term dormant cells are a "protection" against malignancy, because they can survive for long periods and evade treatment, potentially causing relapse many years later.

However, a short-term stem cell that is poised to become active, dividing and reproducing more quickly than a long-term one, can gather up many more mutations, and sometimes these can progress to blood cancers, he adds.

"This research gives us insight into aspects of how cancer starts and how some cancer cells can retain stem-cell like properties that allow them to survive long-term," says Dr. Dick.

He adds that a deeper understanding of stem cells can also help with stem cells transplants for the treatment of blood cancers in the future, by potentially stimulating and growing these cells ex vivo (out of the body) for improved transplantation.

Read the original here:
Mapping out the mystery of blood stem cells - Science Codex

AI makes huge progress predicting how proteins fold one of biology’s greatest challenges promising rapid drug development – The Conversation US

Takeaways

A deep learning software program from Google-owned lab DeepMind showed great progress in solving one of biologys greatest challenges understanding protein folding.

Protein folding is the process by which a protein takes its shape from a string of building blocks to its final three-dimensional structure, which determines its function.

By better predicting how proteins take their structure, or fold, scientists can more quickly develop drugs that, for example, block the action of crucial viral proteins.

Solving what biologists call the protein-folding problem is a big deal. Proteins are the workhorses of cells and are present in all living organisms. They are made up of long chains of amino acids and are vital for the structure of cells and communication between them as well as regulating all of the chemistry in the body.

This week, the Google-owned artificial intelligence company DeepMind demonstrated a deep-learning program called AlphaFold2, which experts are calling a breakthrough toward solving the grand challenge of protein folding.

Proteins are long chains of amino acids linked together like beads on a string. But for a protein to do its job in the cell, it must fold a process of twisting and bending that transforms the molecule into a complex three-dimensional structure that can interact with its target in the cell. If the folding is disrupted, then the protein wont form the correct shape and it wont be able to perform its job inside the body. This can lead to disease as is the case in a common disease like Alzheimers, and rare ones like cystic fibrosis.

Deep learning is a computational technique that uses the often hidden information contained in vast datasets to solve questions of interest. Its been used widely in fields such as games, speech and voice recognition, autonomous cars, science and medicine.

I believe that tools like AlphaFold2 will help scientists to design new types of proteins, ones that may, for example, help break down plastics and fight future viral pandemics and disease.

I am a computational chemist and author of the book The State of Science. My students and I study the structure and properties of fluorescent proteins using protein-folding computer programs based on classical physics.

After decades of study by thousands of research groups, these protein-folding prediction programs are very good at calculating structural changes that occur when we make small alterations to known molecules.

But they havent adequately managed to predict how proteins fold from scratch. Before deep learning came along, the protein-folding problem seemed impossibly hard, and it seemed poised to frustrate computational chemists for many decades to come.

The sequence of the amino acids which is encoded in DNA defines the proteins 3D shape. The shape determines its function. If the structure of the protein changes, it is unable to perform its function. Correctly predicting protein folds based on the amino acid sequence could revolutionize drug design, and explain the causes of new and old diseases.

All proteins with the same sequence of amino acid building blocks fold into the same three-dimensional form, which optimizes the interactions between the amino acids. They do this within milliseconds, although they have an astronomical number of possible configurations available to them about 10 to the power of 300. This massive number is what makes it hard to predict how a protein folds even when scientists know the full sequence of amino acids that go into making it. Previously predicting the structure of protein from the amino acid sequence was impossible. Protein structures were experimentally determined, a time-consuming and expensive endeavor.

Once researchers can better predict how proteins fold, theyll be able to better understand how cells function and how misfolded proteins cause disease. Better protein prediction tools will also help us design drugs that can target a particular topological region of a protein where chemical reactions take place.

The success of DeepMinds protein-folding prediction program, called AlphaFold, is not unexpected. Other deep-learning programs written by DeepMind have demolished the worlds best chess, Go and poker players.

In 2016 Stockfish-8, an open-source chess engine, was the worlds computer chess champion. It evaluated 70 million chess positions per second and had centuries of accumulated human chess strategies and decades of computer experience to draw upon. It played efficiently and brutally, mercilessly beating all its human challengers without an ounce of finesse. Enter deep learning.

On Dec. 7, 2017, Googles deep-learning chess program AlphaZero thrashed Stockfish-8. The chess engines played 100 games, with AlphaZero winning 28 and tying 72. It didnt lose a single game. AlphaZero did only 80,000 calculations per second, as opposed to Stockfish-8s 70 million calculations, and it took just four hours to learn chess from scratch by playing against itself a few million times and optimizing its neural networks as it learned from its experience.

AlphaZero didnt learn anything from humans or chess games played by humans. It taught itself and, in the process, derived strategies never seen before. In a commentary in Science magazine, former world chess champion Garry Kasparov wrote that by learning from playing itself, AlphaZero developed strategies that reflect the truth of chess rather than reflecting the priorities and prejudices of the programmers. Its the embodiment of the clich work smarter, not harder.

Every two years, the worlds top computational chemists test the abilities of their programs to predict the folding of proteins and compete in the Critical Assessment of Structure Prediction (CASP) competition.

In the competition, teams are given the linear sequence of amino acids for about 100 proteins for which the 3D shape is known but hasnt yet been published; they then have to compute how these sequences would fold. In 2018 AlphaFold, the deep-learning rookie at the competition, beat all the traditional programs but barely.

Two years later, on Monday, it was announced that Alphafold2 had won the 2020 competition by a healthy margin. It whipped its competitors, and its predictions were comparable to the existing experimental results determined through gold standard techniques like X-ray diffraction crystallography and cryo-electron microscopy. Soon I expect AlphaFold2 and its progeny will be the methods of choice to determine protein structures before resorting to experimental techniques that require painstaking, laborious work on expensive instrumentation.

One of the reasons for AlphaFold2s success is that it could use the Protein Database, which has over 170,000 experimentally determined 3D structures, to train itself to calculate the correctly folded structures of proteins.

The potential impact of AlphaFold can be appreciated if one compares the number of all published protein structures approximately 170,000 with the 180 million DNA and protein sequences deposited in the Universal Protein Database. AlphaFold will help us sort through treasure troves of DNA sequences hunting for new proteins with unique structures and functions.

As with the chess and Go programs AlphaZero and AlphaGo we dont exactly know what the AlphaFold2 algorithm is doing and why it uses certain correlations, but we do know that it works.

Besides helping us predict the structures of important proteins, understanding AlphaFolds thinking will also help us gain new insights into the mechanism of protein folding.

[Deep knowledge, daily. Sign up for The Conversations newsletter.]

One of the most common fears expressed about AI is that it will lead to large-scale unemployment. AlphaFold still has a significant way to go before it can consistently and successfully predict protein folding.

However, once it has matured and the program can simulate protein folding, computational chemists will be integrally involved in improving the programs, trying to understand the underlying correlations used, and applying the program to solve important problems such as the protein misfolding associated with many diseases such as Alzheimers, Parkinsons, cystic fibrosis and Huntingtons disease.

AlphaFold and its offspring will certainly change the way computational chemists work, but it wont make them redundant. Other areas wont be as fortunate. In the past robots were able to replace humans doing manual labor; with AI, our cognitive skills are also being challenged.

Original post:
AI makes huge progress predicting how proteins fold one of biology's greatest challenges promising rapid drug development - The Conversation US

AI system solves 50-year-old protein folding problem in hours – Livescience.com

An artificial intelligence company that gained fame for designing computer systems that could beat humans at games has now made a huge advancement in biological science.

The company, DeepMind, which is owned by the same parent company as Google, has created an AI system that can rapidly and accurately predict how proteins fold to get their 3D shapes, a surprisingly complex problem that has plagued researchers for decades, according to The New York Times.

Figuring out a protein's structure can require years or even decades of laborious experimentation, and current computer simulations of protein folding fall short on accuracy. But DeepMind's system, known as AlphaFold, required only a few hours to accurately predict a protein's structure, the Times reported.

Related: Why does artificial intelligence scare us so much?

Proteins are large molecules that are essential for life. They are made up of a string of chemical compounds known as amino acids. These "strings" fold in intricate ways to create unique structures that determine what the protein can do. (For example, the "spike" protein on the new coronavirus allows the virus to bind to and invade human cells.)

Nearly 50 years ago, scientists hypothesized that you could predict a protein's structure knowing just its sequence of amino acids. But solving this "protein folding problem" has proved enormously challenging because there are a mind-boggling number of ways in which the same protein could theoretically fold to take on a 3D structure, according to a statement from DeepMind.

Twenty-five years ago, scientists created an international competition to compare various methods of predicting protein structure something of a "protein olympics," known as CASP, which stands for Critical Assessment of Protein Structure Prediction, according to The Guardian.

In this year's challenge, AlphaFold's performance was head and shoulders above its competitors'. It achieved a level of accuracy that researchers were not expecting to see for years.

"This computational work represents a stunning advance on the protein-folding problem, a 50-year-old grand challenge in biology," Venki Ramakrishnan, president of the Royal Society in the United Kingdom, who was not involved with the work, said in a statement. "It has occurred decades before many people in the field would have predicted. It will be exciting to see the many ways in which it will fundamentally change biological research."

For the competition, teams are given the amino acid sequences of about 100 proteins, the structures of which are known but have not been published, according to Nature News. The predictions are given a score from zero to 100, with 90 considered on par with the accuracy of experimental methods.

AlphaFold trained itself to recognize the relationship between the amino acid sequence and protein structure using existing databases. Then, it used a neural network a computer algorithm modeled on the way the human brain processes information to iteratively improve its prediction of the unpublished protein structures.

Overall, AlphaFold had a median score of 92.5. That's up from a score of less than 60 that the system achieved in its first CASP competition in 2018.

The system isn't perfect in particular, AlphaFold did not perform well in modeling groups of proteins that interact with each other, Nature News reported.

But the advance is a game-changer.

"I think it's fair to say this will be very disruptive to the protein-structure-prediction field. I suspect many will leave the field as the core problem has arguably been solved," Mohammed AlQuraishi, a computational biologist at Columbia University told Nature News. "It's a breakthrough of the first order, certainly one of the most significant scientific results of my lifetime."

DeepMind previously made headlines when it created an AI program, known as AlphaGo, that beat humans at the ancient game of Go.

Researchers hope AlphaFold can have many real-world applications. For example, it could help identify the structures of proteins involved in certain diseases and accelerate drug development.

DeepMind is currently working on a peer-reviewed paper on its work on AlphaFold, the Times reported.

Originally published on Live Science.

The rest is here:
AI system solves 50-year-old protein folding problem in hours - Livescience.com

The 3.2- resolution structure of human mTORC2 – Science Advances

RESULTSStructure determination of mTORC2 by cryo-EM

To investigate the structure of mTORC2 and the mechanism of its regulation, we coexpressed recombinant components of human mTORC2 (mTOR, mLST8, Rictor, and SIN1) in Spodoptera frugiperda cells. The assembled complex, purified using tag-directed antibody affinity followed by size exclusion chromatography, was analyzed by cryo-EM (Fig. 1B and figs. S1, A to C, and S2) in the presence of ATPS and either the full-length substrate Akt1 (fig. S1, D and E) or an Akt1 variant missing the PH domain (PH-Akt1), or in the absence of Akt1 with and without ATPS (fig. S2). The sample prepared in the presence of ATPS and PH-Akt1 yielded the highest overall resolution of 3.2 (density A in fig. S2).

mTORC2 forms a rhomboid-shaped dimer (Fig. 1C) as observed in lower-resolution mTORC2 reconstructions (2022). The mTOR kinase consists of the N-terminal Horn and Bridge domains followed by the FAT, FRB, and kinase domains (Fig. 1A). mTOR forms the core of mTORC2 with mLST8 on the periphery, close to the active site cleft, similar to mTOR-mLST8 in mTORC1 (16, 23). In the overall reconstruction, as a consequence of EM refinement of a flexible molecule, one-half of the dimer showed better local resolution (Fig. 1B, fig. S3, A to C, and movie S1). Therefore, focused refinement on a unique half of the assembly improved the resolution to 3.0 (density C in fig. S2), and these maps were used for structural modeling (fig. S3, D to F). Previous mTORC2 and yeast TORC2 reconstructions (2022) revealed that the two mTOR FAT domains are in closer proximity to each other than observed in mTORC1 (16, 23, 27), and in the current structure, the distance between the mTOR FAT domains is further reduced (fig. S3G). Irrespective of these structural differences between the two mTORCs, the catalytic site in mTORC2 closely resembles the catalytic site in mTORC1 without Rheb-mediated activation (23), suggesting that mTORC2 may be activated by a yet to be defined mechanism.

Previous studies of mTORC2 subunits Rictor and SIN1 or their yeast orthologs were not of sufficient resolution to allow de novo model building, resulting in ambiguous or inconsistent interpretations (20, 22, 28). Here, we unambiguously model all structured regions of Rictor and the N-terminal region of SIN1 (Fig. 2, A to C), whereas the middle and C-terminal part of SIN1 retain high flexibility and are not resolved. The fold of Rictor differs substantially from previous interpretations (fig. S3, H and I) (20). Rictor is composed of three interacting stacks of -helical repeats, here referred to as the ARM domain (AD), the HEAT-like domain (HD), and the C-terminal domain (CD) (Fig. 2, A to C). The N-terminal AD (residues 26 to 487) forms a large superhelical arrangement of nine ARM repeats (Fig. 2, A and B) that structurally separates the HD and CD. The HD (residues 526 to 1007), interpreted as two separate domains in previous lower-resolution studies (20, 22), is composed of 10 HEAT-like repeats. In sequence space, the HD and CD of Rictor are separated by an extended stretch of residues (1008 to 1559) that are predicted to be disordered and are not resolved in our reconstruction. We refer to this region as the phosphorylation site region (PR) because it contains most of Rictors phosphorylation sites (29). The two ends of the PR are anchored by a two-stranded -sheet at the top of the HD, which is thus termed the PR anchor (Fig. 2, B and C, and fig. S4A). From here, a partially flexible linker wraps around the AD and the mTOR FRB domain extending toward the CD (Fig. 2B and fig. S4C).

(A) Sequence-level domain organization of Rictor. Flexible and unresolved regions are indicated as dotted lines. Interactions with other proteins in the complex are highlighted below the sequences. Asterisks indicate residues interacting with the N-terminal region of SIN1. (B) Two views of Rictor, colored by domains. The structured part of Rictor forms three domains: an N-terminal Armadillo repeat domain (AD, magenta), a HEAT-like repeat domain (HD, dark magenta), and a C-terminal domain (CD, light red); the phosphorylation site region (PR) remains disordered. The sequences flanking the nonresolved PR are highlighted in red, and the PR anchor is colored in gold. Bound ligands are shown as cyan spheres. (C) Schematic representation of Rictor and SIN1 domain topology. (D) The Rictor CD occupies the FRB domain and sterically blocks FKBP-rapamycin binding (26).

The structured parts of the CD form a four-helix bundle and a zinc finger, with bound Zn2+, in the vicinity of the Rictor N terminus (Fig. 2A and fig. S4, B and C). Residues coordinating the zinc ion are highly conserved in metazoan Rictor (fig. S4F). In earlier work, this domain had been interpreted as representing the SIN1 domain (20). The complete CD is absent in sequences of fungal Rictor orthologs, but other large extensions in yeast Rictor and SIN1 sequences may occupy the equivalent location in yeast TORC2, as observed in an intermediate-resolution reconstruction of budding yeast TORC2 (fig. S4, D and E) (21). Increased levels of Zn2+ have been reported to stimulate Akt1 S473 phosphorylation in cells (30, 31), but no direct involvement of mTORC2 activation has been demonstrated.

Contacts between Rictor and mTOR are made by the Rictor AD, which sits between the Horn domain of the proximal mTOR subunit and the Bridge domain of the distal mTOR subunit (Fig. 2B). With its positioning on top of the mTOR FRB domain, the CD of Rictor blocks the binding space of FKBP12-rapamycin in mTORC1, thereby explaining the absence of an mTORC1-like mode of sensitivity to rapamycin for mTORC2 (Fig. 2D) (5, 8, 28).

The SIN1 subunit of mTORC2 exhibits an unexpected structural organization. The N-terminal region (residues 2 to 137), contrary to earlier interpretations, does not form an independently folding domain but interacts tightly with Rictor and mLST8 in an extended conformation (Figs. 2, A to C, and 3, A to E). The CRIM, Ras-binding domain (RBD), and PH domains of SIN1, however, remain flexibly disposed. The N terminus of SIN1 is inserted into a deep cleft at the interface of the AD and HD of Rictor. The N-terminal Ala2 with a structurally resolved acetylated N terminus and Phe3 of SIN1 are buried in a hydrophobic pocket of Rictor (Fig. 3, C and D, and fig. S5A). The anchored N-terminal region of SIN1 forms two short helices (residues 6 to 33) inserted into grooves on the surface of the Rictor AD (Fig. 3D) and then continues with a flexible sequence segment toward the Rictor CD (Figs. 2, B and C, and 3C and fig. S5B). Protruding from the Rictor CD, SIN1 forms a helical segment, referred to as the traverse, that spans the distance to mLST8 across the mTORC2 kinase cleft (Fig. 3C and fig. S5, B and C). The next region of SIN1 interacts with the fourth strand of the second blade of the mLST8 propeller by -strand complementation, leading to displacement of an mLST8 loop relative to the structure of mLST8 in mTORC1 (Fig. 3, C and E, and fig. S5D). SIN1 then follows the surface of the mLST8 propeller, finally forming an -helix anchored between the first and seventh blades of mLST8.

(A) Sequence-level domain organization of SIN1. Flexible and unresolved regions are shown above each domain representation as dotted lines in two colors as indicated. Interactions with other proteins in the complex are indicated below the domain representation. (B) Extension of the processed SIN1 N terminus disrupts assembly of Rictor and SIN1 with mTOR/mLST8 into mTORC2. SDS-polyacrylamide gel of a FLAG bead pulldown from lysates of insect cells expressing mTORC2 comprising SIN1 variants. Levels of Rictor are drastically reduced in the mTOR-based pulldown for mTORC2 carrying variants of SIN1 N-terminally extended by a tryptophan (mTORC2 SIN1_W), two consecutive arginines (mTORC2 SIN1_2R), and three consecutive arginines (mTORC2 SIN1_3R). (C) Surface representation of mTORC2. SIN1 (shown as green cartoon) interacts via two N-terminal helices with Rictor, winds around Rictor, traverses the catalytic site cleft, and winds around mLST8. The field of view of subpanel D is indicated. (D) Close-up view of the SIN1 N-terminal residues, which are deeply inserted between Rictor AD and HD. Acetylated Ala2 and Phe3 are bound in a hydrophobic pocket, while Asp5 interacts via salt bridges (yellow dashes). (E) Top view of mLST8 -propeller (orange) and the interaction regions with SIN1 (green). The nomenclature for WD40 -propeller repeats is indicated. (F) Top view of the catalytic site with the structure shown as surface together with the density of a subclass (light gray). The lower-resolution extra density is consistent with a placement of the SIN1 CRIM domain, here shown in dark green (PDB: 2RVK). Unassigned extra density protrudes from the CRIM domain to the mTOR active site and Rictor.

SIN1 integrates into the Rictor fold and connects Rictor with mLST8, suggesting a direct role in stabilizing mTORC2. To test the relevance of the anchoring of the N terminus of SIN1 on Rictor, we extended the N terminus of SIN1 using tryptophan or arginine residues to exploit steric hindrance or charge-charge repulsion to prevent the insertion into the Rictor pocket. Insertion of residues impairs critical interactions observed for the acetylated N terminus of SIN1 and prevents Rictor integration into mTORC2, as observed in baculovirus-mediated expression of mTOR components followed by pull-down assays (Fig. 3B and fig. S5E). Therefore, SIN1 acts as an integral part of the Rictor structure that critically stabilizes interdomain interactions, explaining the difficulties observed in purifying isolated Rictor (20).

These observations are also consistent with the locations of posttranslational modifications or mutations that affect mTORC2 activity. SIN1 phosphorylation at Thr86 and Thr398 has been reported to reduce mTORC2 integrity and kinase activity toward Akt1 Ser473 (32). Thr86 in SIN1, which is a target for phosphorylation by S6 kinase (32), is bound to a negatively charged pocket of the Rictor CD (Fig. 3C and fig. S5C). Phosphorylation of Thr86 would lead to repulsion from this pocket, destabilizing the interaction between Rictor and mTOR-mLST8 and presumably the entire mTORC2 assembly, in agreement with earlier in vivo and in vitro observations (32). The importance of SIN1 in connecting Rictor to mLST8, and therefore also indirectly to mTOR, is also consistent with the requirement of mLST8 for mTORC2 integrity (33, 34).

A poorly resolved density linked to the SIN1 helix anchored to mLST8 is observed in all reconstructions. In previous structural studies of yeast TORC2, a similar region of density was associated with the CRIM domain of Avo1, the yeast SIN1 ortholog (21, 28). Most likely, it represents the mobile substrate-binding CRIM domain that directly follows the helix in the SIN1 sequence and has a matching shape based on the solution structure of the Schizosaccharomyces pombe SIN1 CRIM domain (Fig. 3F and fig. S6, A to C) (25, 26). The positions of the SIN1 RBD and PH domains remain unresolved. In the dataset collected for samples with added full-length Akt1 (dataset 2 in fig. S2), we observed additional low-resolution density (Fig. 3F and fig. S6, B and C) between the hypothetic CRIM domain and Rictor AD and CD in the vicinity of the mTOR active site. This density, not of sufficient resolution to assign specific interactions, may represent parts of bound Akt1 or SIN1 domains (fig. S6C).

A proposed regulatory mechanism for mTORC2 involves ubiquitylation of mLST8 on Lys305 and Lys313 (35). Loss of ubiquitylation by K305R and/or K313R mutation, or truncation of mLST8 at Tyr297, leads to mTORC2 hyperactivation and increased AKT phosphorylation (35). mLST8 Lys305 is proximal to the SIN1 helix anchoring the CRIM domain. Ubiquitylation of Lys305 would prevent association of the SIN1 helix, leading to dislocation of the SIN1 CRIM domain required for substrate recruitment (Figs. 3C and 4A). Ubiquitylation of Lys313, which is found on the lower face of mLST8 (Figs. 3C and 4A), presumably also interferes with positioning of the CRIM domain (fig. S6A).

(A) Overview of mTORC2 architecture and ligand interaction sites. Each half of the dimeric mTORC2 has three small-molecule binding sites. The kinase active site and the A-site, which is located in the peripheral region of Rictor, bind to ATP (or ATP analogs). The I-site in the middle of the FAT domain of mTOR binds InsP6. The indicated modifications on SIN1 and mLST8 affect mTORC2 assembly. Extra-density region following the CRIM domain is indicated as a gray outline. (B). Close-up view of the A-site on the periphery of the Rictor HD with bound ATPS. A hydrogen bond between ATPS and Asn543 is shown as dashed yellow lines. (C) Close-up view of the I-site in the FAT domain of mTOR. InsP6 is surrounded by a cluster of positively charged amino acids. It only directly interacts with residues of the FAT domain.

We observed two previously uncharacterized, small-molecule binding sites outside the mTOR catalytic site, which is itself occupied by ATPS. The first (A-site) (Fig. 4B and fig. S7, A and B) is located in the HD of Rictor and is thus specific to mTORC2. The second (I-site) (Fig. 4C and fig. S7C) is located in the FAT domain of mTOR and is thus common to mTORC1 and mTORC2.

The density of the small molecule in the A-site matched that of an ATP molecule and was confirmed to be ATP (or ATPS) through a comparison of cryo-EM reconstructions of mTORC2 with and without ATPS added at a near physiological concentration of 2 mM (datasets 1 and 4, figs. S2 and S7A). The A-site does not resemble any known ATP-binding site. Positively charged amino acids (Lys541, Arg575, Arg576, and Arg572) of the A-site are conserved in Rictor orthologs from yeast to human (figs. S4E and S8). Other residues are not conserved, hinting at the possibility for interactions with alternative negatively charged ligands. The A-site is located approximately 100 from the mTOR catalytic site. Ligand binding to the A-site caused neither long-range allosteric change affecting the kinase site nor local structural perturbations (fig. S9, I to L).

To investigate the effect of ligand binding to the A-site, we generated a series of Rictor variants with a mutated A-site (table S1). Variants with three or four mutated residues (A3 and A4) assembled into mTORC2 (fig. S10B), while variant A5 was defective in assembly (fig. S10, B to D). Cryo-EM reconstructions of variants A3 and A4 in the presence of ATPS (fig. S9, I to L) confirmed that the chosen mutations abolish ligand binding under near physiological conditions (figs. S7A and S9, J and L). Purified mTORC2 containing Rictor variant A3 or A4 exhibited thermal stability and kinase activity, in an Akt1 in vitro phosphorylation assay, comparable to wild-type (WT) mTORC2 (fig. S10, F to H). Complementation of a Rictor knockout (KO) in human embryonic kidney (HEK) 293T cells by transfected Rictor-WT, or Rictor variant A3 yielded comparable levels of Akt1-S473 phosphorylation (table S1 and fig. S11). Together, the above analyses indicate that ligand binding to the A-site does not directly influence mTORC2 kinase activity, suggesting rather a role in the interaction with other, yet unidentified, partner proteins of mTORC2.

The I-site is formed entirely by the FAT domain of mTOR, where a large, positively charged, pocket is lined by six lysine and two arginine residues to bind an extended ligand (Fig. 4C and fig. S7C). The I-site was still partially occupied in our reconstruction of mTORC2 prepared without addition of exogenous ATPS or other relevant ligands (fig. S7A). The copurified molecule was identified by map appearance and by ion mobility spectrometrymass spectrometry (IMS-MS) as inositol hexakisphosphate (InsP6) (figs. S7, D to F, and S12). InsP6 binds in a region, which is incomplete in related PI3Ks (36), but is generally conserved in members of the PIKK family of kinases (37). InsP6 was previously reported to associate with DNA-PKcs (38). Recently, structure determination of the PIKK family kinase SMG1 revealed InsP6 binding in a region corresponding to the I-site and led the authors to postulate a corresponding binding site in mTOR but involving both the kinase domain and FAT domain (37). InsP6 has previously been observed as a structural component of multi-subunit assemblies, including the spliceosome (39) and proteasome activator complex (40), and helical repeat regions have been identified as InsP6 interaction sites (41).

To investigate the function of InsP6 interaction, we purified recombinant mTORC2 containing mTOR I-site mutations (table S1). mTOR variants with two and three mutations, I2 and I3, yielded intact mTORC2 complexes (fig. S10A), while a variant with five mutations, I5, failed to assemble into mTORC2 (fig. S10, A and D). mTORC2 containing mTOR variants I2 and I3 displayed normal kinase activity toward Akt1 in vitro (fig. S10E). Notably, the mutations in I2 are equivalent to those reported previously to abolish completely the kinase activity of an N-terminally truncated naked mTOR fragment toward a C-terminal peptide of Akt1 (37). A possible explanation for this apparent discrepancy is provided by a reduced stability of mTORC2 assembled using the I2 variant (but not the I3 variant) (fig. S10G). This destabilizing effect might be more pronounced in an mTOR fragment than in the context of an assembled mTORC2 (fig. S10G).

To investigate a possible role of InsP6 metabolism on mTORC2 activity in HEK293T cells, we knocked down (KD) and knocked out (KO) inositol-pentakisphosphate 2-kinase (IPPK) and multiple inositol polyphosphate phosphatase 1 (MINPP1), respectively. The former enzyme generates InsP6, whereas the latter degrades it (fig. S13). These manipulations of InsP6-metabolizing enzymes did not alter mTORC2 kinase activity in nonstimulated cells or in cells stimulated with fetal calf serum (FCS) and insulin (fig. S13). These biochemical results are consistent with the observed stable binding of InsP6 to mTORC2 and suggest a role of InsP6 in mTOR folding or mTOR complex assembly, rather than as an acute transient metabolic input signal to mTORC1 or mTORC2.

Insect cell vectors from the MultiBac Baculovirus expression system (42) (Geneva Biotech, Geneva, Switzerland) have been used to clone internally FLAG-tagged pAceBAC-mTOR (FLAG after Asp258), pIDK-Rictor, pIDC-mLST8, and pAceBAC1-SIN1 using Gateway Cloning (Thermo Fisher Scientific, USA). Rictor was originally amplified from myc-Rictor, which was a gift from D. Sabatini (8) (Addgene plasmid no. 11367). Site-directed mutagenesis was used to generate mTORC2 A- and I-site variants. The following set of A-site mutants with pIDK-Rictor as template was created: Rictor_R572E_R575E_R576E (A3), Rictor_R572E_R575E_R576E_Y579A (A4), and Rictor_R572E_R575E_R576E_Y579A_L587W (A5). The following I-site mutants with FLAG-tagged pAceBAC-mTOR were generated: mTOR_K1753E_K1788E (I2), mTOR_R1628E_K1655E_K1662E (I3), and mTOR_R1628E_K1655E_K1662E_K1706E_K1735E (I5). WT Rictor and mutants A3 and A5 were subcloned into plasmid MX01 (Addgene plasmid no. 158624). SIN1 N-terminal variants were generated by inserting a tryptophan (SIN1_W), two consecutive arginines (SIN1_2R), or three consecutive arginines (SIN1_3R) using site-directed mutagenesis and pAceBAC1-SIN1 as template. Plasmids encoding FLAG-tagged mTOR, Rictor, and mLST8 were fused to a MultiBac expression plasmid using Cre-recombinase (New England Biolabs, Ipswich, USA) and transposed into a bacmid for baculovirus production. Baculovirus encoding untagged SIN1 was produced separately.

Sf21 insect cells (Expression Systems) were grown in HyClone insect cell media (GE Life Sciences), and baculovirus was generated according to Fitzgerald et al. (42). For the expression of recombinant human WT mTORC2, A- and I-site mTORC2 mutants, and mTORC2 carrying SIN1 N-terminal variants, Sf21 cells were infected at a cell density of 1 Mio/ml. Cells were coinfected with 1:100 (v/v) ratio of two undiluted supernatants from cells previously infected with baculovirus encoding FLAG-mTOR, Rictor, and mLST8 or infected with baculovirus encoding untagged SIN1, respectively. WT mTORC2, A-site mutants A3, A4, and A5, and I-site mutants I2, I3, and I5 were purified as follows: Insect cells were harvested 72 hours after infection by centrifugation at 800g for 25 min and stored at 80C until further use. Cell pellets were lysed in 50 mM bicine (pH 8.5), 200 mM NaCl, and 2 mM MgCl2 by sonication, and the lysate was cleared by ultracentrifugation. Soluble protein was incubated with 10 ml of anti-DYKDDDDK agarose beads (Genscript, Piscataway, USA) for 1 hour at 4C. The beads were transferred to a 50-ml gravity flow column (Bio-Rad) and washed four times with 200 ml of wash buffer containing 50 mM bicine (pH 8.5), 200 mM NaCl, and 2 mM EDTA. Protein was eluted by incubating beads for 30 min with 10 ml of wash buffer supplemented with synthetic DYKDDDDK peptide (0.6 mg/ml) (Genscript, Piscataway, USA). The eluate was combined with three additional elution steps using synthetic DYKDDDDK peptide (0.1 mg/ml) and 5-min incubation time. The eluted protein was concentrated using a 100,000-Da molecular mass cutoff centrifugal concentrator (Amicon) of regenerated cellulose membrane and purified by size exclusion chromatography on a custom-made Superose 6 Increase 10/600 GL gel filtration column equilibrated with 10 mM bicine (pH 8.5), 150 mM NaCl, 0.5 mM EDTA, and 2 mM tris(2-carboxyethyl)phosphine (TCEP). Purified WT mTORC2 was concentrated in gel filtration buffer to a final concentration of 3 to 3.5 mg/ml determined by A280 absorption using NanoDrop 2000 (Thermo Fisher Scientific). Sample was supplemented with 5% (v/v) glycerol and stored at 80C for later cryo-EM use. Purified mTORC2 variants with A- and I-site mutants were concentrated in gel filtration buffer to a final concentration of 0.4 to 2 mg/ml as determined by absorption at 280-nm wavelength using NanoDrop 2000 (Thermo Fisher Scientific). The resulting samples were supplemented with 5% (v/v) glycerol and stored at 80C for later use.

The coding sequence for Akt1 (43) was cloned into a pAceBAC1 expression vector (Geneva Biotech, Geneva, Switzerland) with an N-terminal His10-Myc-FLAG tag by Gateway cloning. Baculovirus was produced as described for mTORC2. Akt1 was purified with anti-DYKDDDDK agarose beads as described for mTORC2. The eluted protein was concentrated using a 10,000-Da molecular mass cutoff centrifugal concentrator (Amicon) of regenerated cellulose membrane and further purified by size exclusion chromatography with a Superdex 75 Increase column equilibrated with 10 mM bicine (pH 8.5), 150 mM NaCl, 0.5 mM EDTA, and 2 mM TCEP. Purified Akt1 was concentrated in gel filtration buffer, supplemented with 5% (v/v) glycerol, and stored at 80C for further experiments. Dephosphorylated Akt1 was obtained after overnight incubation of 4.5 mg of protein with 6 g of -protein phosphatase (New England Biolabs) in the presence of PMP buffer (New England Biolabs) and 1 mM MnCl2 before size exclusion chromatography. Successful Akt1 dephosphorylation was confirmed by Western blot with antibodies against phosphoAKT-Ser473 (no. 4060; Cell Signaling Technology, Beverly, USA) and phosphoAKT-Thr450 (no. 9267; Cell Signaling Technology, Beverly, USA) at a dilution of 1:1000 in 5 ml of Tris-buffered saline with 0.1% Tween20 (TBST). Human (Delta-PH) Akt1 protein (residues 144 to 480, mono-phosphorylated on T450), as described by Lui et al. (44) (therein referred to as Akt1KD), was provided by T. Leonard (Max-Perutz Labs, Vienna).

A-site mutants A3, A4, and A5 and I-site mutants I2, I3, and I5, and mTORC2 carrying SIN1 N-terminal variants extended by a tryptophan (SIN1_W), two consecutive arginines (SIN1_2R), and three consecutive arginines (SIN1_3R) inserted between the processed Met1 and Ala2, were immunoprecipitated in small scale using FLAG beads. Five-gram wet weight of pellets from insect cells expressing A- and I-site mutants and SIN1 N-terminal variants was lysed in 50 mM bicine (pH 8.5), 200 mM NaCl, and 2 mM MgCl2 using a Dounce homogenizer. The lysate was cleared by ultracentrifugation for 45 min at 35,000g. Soluble protein was incubated with 125 l of anti-DYKDDDDK agarose beads (Genscript, Piscataway, USA) for 1 hour at 4C. The beads were transferred to a 5-ml gravity flow column (Pierce Centrifuge Columns, Thermo Fisher Scientific) and washed with 50 ml of buffer containing 50 mM bicine (pH 8.5), 200 mM NaCl, and 2 mM EDTA. Protein was eluted by 30-min incubation of the beads with 400-l wash buffer supplemented with synthetic DYKDDDDK peptide (0.6 mg/ml) (Genscript, Piscataway, USA). Total lysate, soluble supernatant after ultracentrifugation, flow through from FLAG column, buffer wash, and elution fraction were loaded onto a 4 to 15% SDS polyacrylamide gel (Bio-Rad Laboratories). In addition, total lysate, supernatant after ultracentrifugation, and elution fraction of mTORC2 WT, SIN1 N-terminal variants, and mutants A5 and I5 were analyzed by immunoblotting using antibodies against mTOR (no. 2972; Cell Signaling Technology, Beverly, USA), SIN1 (A300-910A; Bethyl), Rictor (A300-458A; Bethyl), and actin (MAB1501; Merck Millipore) at a dilution of 1:1000 in 5 ml of TBST. A goat anti-rabbit horseradish peroxidase (HRP)labeled antibody (ab6721; Abcam, Cambridge, UK) was used as the secondary antibody at a dilution of 1:3000 in 5 ml of TBST.

mTORC2 kinase activity assays were conducted in 100 mM Hepes (pH 7.4), 1 mM EGTA, 1 mM TCEP, 0.0025% Tween 20, and 10 mM MnCl2 using dephosphorylated Akt1 as a substrate. In a 60-l reaction volume, 0.05 M of either WT or A- and I-site mutant mTORC2 was mixed with 1 M Akt1 and, where indicated, either dimethyl sulfoxide or 25 M Torin1. The mixture was preincubated for 5 min at room temperature, and the reaction was initiated by the addition of 10 M ATP. After 20 min at 37C, the reaction was terminated by the addition of 60 l of 2 Laemmli sample buffer. The reactions were analyzed by Western blotting using primary antibodies against phosphoAKT-Ser473 (no. 4060; Cell Signaling Technology, Beverly, USA), phosphoAKT-Thr450 (no. 9267; Cell Signaling Technology, Beverly, USA), AKT (no. 4685), and mTOR (no. 2972; Cell Signaling Technology, Beverly, USA), anti-FLAG antibodies (Sigma-Aldrich, F1804), SIN1 (Bethyl, A300-910A), and Rictor (Bethyl, A300-458A) at a dilution of 1:1000 in 5 ml of TBST. A goat anti-rabbit HRP-labeled antibody (ab6721; Abcam, Cambridge, UK) was used as the secondary antibody at a dilution of 1:3000 in 5 ml of TBST. Signals were detected using the Enhanced Chemiluminescence (ECL) Kit SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific). Images were acquired using a Fusion FX (Vilber) imaging system.

Thermal unfolding was monitored by differential scanning fluorimetry (DSF) based on internal tryptophane fluorescence on a Prometheus NT.48 instrument (NanoTemper Technologies). Purified WT mTORC2 or mTORC2 containing mutations in A- or I-site was diluted to 0.1 mg/ml in 10 mM bicine (pH 8.5), 150 mM NaCl, 0.5 mM EDTA, and 2 mM TCEP. High-precision capillaries (NanoTemper Technologies) were filled with 10-l sample and placed on the sample holder. A temperature gradient of 0.1C/min from 22 to 65C was applied, and fluorescence intensity at 330 and 350 nm was recorded. A plot of the ratio of fluorescence intensities at those wavelengths (F350/F330) was generated using a Python script. The experiment was repeated two times with five replicates per sample run each time. Melting points were calculated using PR.ThermControl software version 2.1.2. Data were analyzed using GraphPad Prism version 8.0.0 (GraphPad Software, San Diego, CA, USA) to generate the mean and SD of the melting points. One outlier, likely resulting from capillary handling, for sample A4 was excluded from data analysis.

HEK293T cells were cultured and maintained in Dulbeccos modified Eagles medium (DMEM) high glucose with 10% FCS, 4 mM glutamine, 1 mM sodium pyruvate, and 1 penicillin/streptomycin. RICTOR KO cells were generated as described by Bossler et al. (45). Four micrograms of plasmids harboring RICTOR-WT, RICTOR-A_3, and RICTOR-A_5 was transfected with jetPRIME (Polyplus). Twenty-four hours after transfection, cells were starved for serum for overnight and stimulated with 10% FCS and 100 nM insulin for 15 min. Total cell lysates were prepared in lysis buffer containing 100 mM tris-HCl (pH 7.5), 2 mM EDTA, 2 mM EGTA, 150 mM NaCl, 1% Triton X-100, complete inhibitor cocktail (Roche), and PhosSTOP (Roche). Protein concentration was determined by a Bradford assay, and equal amounts of protein were separated by SDSpolyacrylamide gel electrophoresis (SDS-PAGE) and transferred onto nitrocellulose membranes (GE Healthcare). Antibodies used were as follows: AKT (1:1000 dilution, catalog no. 2920, Cell Signaling Technology), AKT-pS473 (1:1000, catalog no. 4060, Cell Signaling Technology), RICTOR (1:1000, catalog no. 2040, Cell Signaling Technology), ACTIN (1:2000, catalog no. MAB1501, Millipore), IRDye 800CW goat anti-rabbit immunoglobulin G (IgG) (1:20,000, catalog no. 926-32211, LI-COR), and IRDye 680RD goat anti-mouse IgG (1:20,000, catalog no. 926-68070). All antibodies were diluted in 10 ml of TBST and Licor intercept (TBS) blocking buffer (1:1). Signals were detected by LI-COR Fc (LI-COR Biosciences).

HEK293T cells were cultured and maintained in DMEM high glucose with 10% FCS, 4 mM glutamine, 1 mM sodium pyruvate, and 1 penicillin/streptomycin. For KD of IPPK and MINPP1, 0.1 106 cells per well were seeded in a six-well plate and transfected with 100 nM small interfering RNA (siRNA) using the jetPRIME (Polyplus) system. After 32 hours, cells were washed twice with phosphate-buffered saline (PBS) (/) and starved for serum for 16 hours. Forty-eight hours after transfection, cells were incubated at 37C with PBS (+/+) for 10 min followed by stimulation with 10% FCS and 100 nM insulin for 15 min at 37C. Cells were washed with ice-cold PBS (/) and harvested for SDS-PAGE or RNA isolation for quantitative polymerase chain reaction (qPCR) analysis. KO experiments were conducted as described above, using generated KO cells instead of transfection with siRNA. Total cell lysates were prepared in M-PER lysis buffer (Thermo Fisher Scientific) containing complete inhibitor cocktail (Roche) and PhosSTOP (Roche), and protein concentrations were determined by Bradford assay. Equal amounts of protein were separated by SDS-PAGE and transferred onto nitrocellulose membranes (GE Healthcare), and signals were detected by LI-COR Fc (LI-COR Biosciences). Antibodies used were as follows: AKT (1:1000, catalog no. 2920, Cell Signaling Technology), AKT-pS473 (1:1000, catalog no. 4060, Cell Signaling Technology), ACTIN (1:5000, catalog no. MAB1501, Millipore), IRDye 800CW goat anti-rabbit IgG (1:20,000, catalog no. 926-32211, LI-COR), and IRDye 680RD goat anti-mouse IgG (1:20,000, catalog no. 926-68070). All antibodies were diluted in 10 ml of TBST and Licor intercept (TBS) blocking buffer (1:1).

For qPCR, total RNA was isolated using the RNeasy Kit (Qiagen). RNA was reverse-transcribed to complementary DNA (cDNA) using the iScript cDNA Synthesis Kit (Bio-Rad). Semiquantitative real-time PCR analysis was performed using Fast SYBR Green (Applied Biosystems). Relative expression levels were determined by normalizing each CT values to POLR2A using the CT method. The sequence for the primers used in this study was as follows: IPPK-fw, 5-AATGAATGGGGGTACCACGG-3; IPPK-rv, 5-AACTTCAGAAACCGCAGCAC-3; MINPP1-fw, 5-AGCTACTTTGCAAGTGCCAG-3; MINPP1-rv, 5-TGCATGACCAAACTGGAGGA-3.

KO cells were generated using the LentiCRISPR system as described by Sanjana et al. (46). Guide RNAs (gRNAs) against IPPK and MINPP1 were expressed from LentiCRISPRv2 (gifts from F. Zhang; Addgene plasmid nos. 49535 and 52961) by transfection of HEK293T cells with 1 g of DNA using jetPRIME. The following gRNA target sequences were used: IPPK gRNA, 5-TCGGCCGGTGCTCTGCAAAG-3; MINPP1 gRNA, 5-ATCCAGTCCGCGTACCACAA-3. Following transfection, cells were selected with puromycin, propagated, and screened for loss of target protein by qPCR. DNA sequencing of PCR products confirmed insertions or deletions leading to interrupted sequencing reactions. Pools of KO cells were used to avoid clonal variation. HEK293T cells transfected with empty vector were used as control.

Ten micrograms of mTORC2 I-site mutants I2 and I3 and A-site mutants A3, A4, and A5 was dissolved in 50 l of digestion buffer [1% sodium deoxycholate (SDC), 0.1 M tris, 10 mM TCEP, 15 mM chloroacetamide (CAA) (pH 8.5)] using vortexing for trypsin digestion. For endoproteinase GluC and chymotrypsin digestion, the same protein aliquots were dissolved in 20 l of a digestion buffer consisting of 1 M urea, 0.1 M ammonium bicarbonate, 10 mM TCEP, and 15 mM CAA. Samples were either incubated for 10 min at 95C (trypsin) or 1 hour at 37C (GluC and chymotrypsin) to reduce and alkylate disulfide bonds. Protein aliquots were digested overnight at 37C by incubation with sequencing-grade modified trypsin, GluC, and chymotrypsin (all 1:50, w/w; Promega). Then, the peptides were cleaned up using iST cartridges (PreOmics, Munich) according to the manufacturers instructions. Samples were dried under vacuum and dissolved in LC-buffer A (0.1% formic acid) at a concentration of 0.05 g/l.

To enhance the sensitivity of the liquid chromatographyMS (LC-MS) analysis, a label-free targeted LC-MS approach was carried out. Therefore, three lists of peptides considering the cleavage specificity of the three proteases used and containing all mutation sites were generated. The peptide sequences were imported into Skyline (version 20.1; https://brendanx-uw1.gs.washington.edu/labkey/project/home/software/Skyline/begin.view) to generate a mass isolation list of all doubly and triply charged precursor ions for each protease. These were then loaded into a Q Exactive plus LC-MS platform and analyzed using the following settings: The setup of the RPLC-MS system was as described previously (47). Chromatographic separation of peptides was carried out using an EASY nano-LC 1000 system (Thermo Fisher Scientific), equipped with a heated RP-HPLC column (75 m by 30 cm) packed in-house with 1.9-m C18 resin (Reprosil-AQ Pur, Maisch). Peptides were analyzed per LC-MS/MS run using a linear gradient ranging from 95% solvent A (0.15% formic acid and 2% acetonitrile) and 5% solvent B (98% acetonitrile, 2% water, and 0.15% formic acid) to 45% solvent B over 60 min at a flow rate of 200 nl/min. MS analysis was performed on a Q Exactive plus mass spectrometer equipped with a nano-electrospray ion source (both Thermo Fisher Scientific). Each MS cycle consisted of one MS1 scan followed by high-collision dissociation of the selected precursor ions in the isolation mass lists. Total cycle time was approximately 2 s. For MS1, 3 106 ions were accumulated in the Orbitrap cell over a maximum time of 50 ms and scanned at a resolution of 35,000 FWHM [at 200 mass/charge ratio (m/z)]. MS2 scans were acquired at a target setting of 3 106 ions, accumulation time of 110 ms, and a resolution of 35,000 FWHM (at 200 m/z). The normalized collision energy was set to 27%, the mass isolation window was set to 0.4 m/z, and one microscan was acquired for each spectrum.

The acquired raw files were converted to the mascot generic file (mgf) format using the msconvert tool [part of ProteoWizard, version 3.0.4624 (2013-6-3)]. Using the MASCOT algorithm (Matrix Science, version 2.4.1), the mgf files were searched against a decoy database containing normal and reverse sequences of the predicted SwissProt entries of Homo sapiens (www.ebi.ac.uk, release date 9 December 2019), the mTOR and Rictor mutations, and commonly observed contaminants (in total 41,556 sequences for H. sapiens) generated using the SequenceReverser tool from the MaxQuant software (version 1.0.13.13). The precursor ion tolerance was set to 10 ppm, and fragment ion tolerance was set to 0.02 Da. The search criteria were set as follows: Full tryptic specificity was required (cleavage after lysine or arginine residues unless followed by proline), three missed cleavages were allowed, carbamidomethylation (C) was set as a fixed modification, and oxidation (M) was set as a variable modification. Next, the database search results were imported to the Scaffold Q+ software (version 4.3.2, Proteome Software Inc., Portland, OR), and the protein false identification rate was set to 1% based on the number of decoy hits. Specifically, peptide identifications were accepted if they could be established at greater than 97.0% probability to achieve a false discovery rate less than 1.0% by the scaffold local FDR algorithm. Protein identifications were accepted if they could be established at greater than 65.0% probability to achieve an FDR less than 1.0% and contained at least one identified peptide. Protein probabilities were assigned by the Protein Prophet program (48). Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony. Proteins sharing significant peptide evidence were grouped into clusters. Last, a spectral library (*.blib) was generated from the assigned MS/MS spectra and imported to Skyline together with the acquired raw data files. Only precursor ions confidently identified by database searching and present in the spectral library were used for quantitative analysis. Quantitative result reports were further analyzed by Microsoft Excel and PRISM (GraphPad Software, San Diego, USA).

Different conditions were screened for mTORC2 in the presence and absence of substrates (fig. S2). For all conditions, freshly thawed mTORC2 aliquots were used to prepare samples with an mTORC2 concentration of 0.37 mg/ml. Shortly before grid preparation, the samples were diluted to reach a final mTORC2 concentration of 0.12 mg/ml.

For each grid, a small piece of continuous carbon was floated on top of the sample for 1 min. The carbon was then picked with a Quantifoil R2/2 holey carbon copper grid (Quantifoil Micro Tools), which was swiftly mounted in a Vitrobot (Thermo Fischer Scientific) whose chamber was set to 4C and 100% humidity. Five microliters of buffer was then added on top of the grid on the side showing the carbon covered with particles, which was immediately blotted with a setting of 0- to 6-s blotting time and rapidly plunge-frozen in a mixture 2:1 of propane:ethane (Carbagas).

Data were collected using a Titan Krios (Thermo Fisher Scientific) transmission electron microscope equipped with either a K2 Summit direct electron detector (Gatan), a K3 direct electron detector (Gatan), or a Falcon 3EC direct electron detector (Thermo Fisher Scientific) using either EPU (Thermo Fisher Scientific) or SerialEM (fig. S2) (49). Cameras were used in counting and/or super-resolution mode. During data collection, the defocus was varied between 1 and 3 m and four exposures were collected per holes. Stacks of frames were collected with a pixel size of 0.84 /pixel and a total dose of about 70 e/2.

For all datasets, the initial processing was done in similar fashion. First, the stacks of frames were aligned and dose-weighted using Motioncor2 (50). GCTF (51) was used to estimate the contrast transfer function (CTF) of the nondose-weighted micrographs. After a selection of good micrographs using both the quality of the power spectra and the quality of the micrographs themselves as criteria, particles were picked using batchboxer from the EMAN1.9 package (52) using particle averages from manually picked particles as references. Particles were extracted using Relion3.0 (53), followed by two rounds of two-dimensional (2D) classification using cryoSPARCv2 (Structura Biotechnology Inc.) (table S2) (54). The first reference was generated by ab initio reconstruction using cryoSPARCv2. Good particles from 2D classification were then used for a homogeneous 3D refinement followed by nonuniform refinement using cryoSPARCv2. Two masks were then generated manually around each half of the pseudo-dimeric mTORC2 using UCSF Chimera (55), and two focused refinements around each half of the complex using cryoSPARCv2 were performed using those masks. For dataset 1, which contained PH-Akt1, the resolution was further improved by performing Bayesian particle polishing (53) followed by CTF refinement using Relion3.1. Those particles were again subjected to a round of nonuniform refinement and local refinement using cryoSPARC v2. For each reconstruction, the maps were sharpened using phenix.auto_sharpen (56) or were transformed to structure factors using phenix.map_to_structure_factors (56) and sharpened in COOT (57).

Further 3D classifications without alignment for local structural variability close to the catalytic center were performed using the particles from the datasets containing the purified Akt1 and, independently, the ones from the dataset with PH-Akt1 using Relion3.0 (53) and using a mask manually created in UCSF Chimera (55). After classification, the particles were used for refinement using cryoSPARCv2 (Structura Biotechnology Inc.). To compare the density of the sample with and without ATPS, the final density (volume A) was filtered to 4.2 and compared to the density without ATPS (volume F). Difference density was calculated using UCSF ChimeraX (58).

First, mTOR and mLST8 models were taken from the EM structure of mTORC2 [Protein Data Bank (PDB): 5ZCS (20)] and each fold was rigid bodyfitted into the better half of the density. Minor changes in mTOR conformation were done manually to fit the density, and then Rictor and SIN1 were manually built de novo using COOT (57). Map quality enabled direct model building for structured regions, and lower-resolution density provided connectivity information for assigning and linking regions of Sin1 and Rictor as shown in figs. S4C and S5B. The second half of mTORC2 was made by copying and rigid body fitting each chain of the first half in the second one. Last, the structure of either one- or two-sided mTORC2 was refined using phenix.real_space_refine (table S2) (56), using Ramachandran and secondary structure restraints. As the horns of mTOR were flexible and their local resolutions were considerably lower, additional reference restraints were applied, using PDB: 6BCX (23) as reference. The model was then validated by comparing the Fourier Shell Correlations (FSC) calculated for the experimental density and the models (fig. S3). In addition, both the half and full structure were also refined in their respective half map (half map 1) and the FSCs of this structure against the same half map (half map 1), the other half (half map 2), and the full map were compared. The similarity of the curves shows that the structure was not overfitted.

InsP6 (Sigma-Aldrich) was directly dissolved in 10 mM ammonium acetate (pH 8.5) and diluted to 50 M. mTORC2 in cryo-EM buffer was buffer-exchanged and concentrated in 10 mM ammonium acetate (pH 8.5) using an Amicon Ultra-0.5 mLMWCO 100kDa. The concentrated complex was mixed with an equal volume of Phenol at pH 8, thoroughly vortexed for 30 s, and incubated at room temperature for 30 min. The tube was then centrifuged for 5 min at 15,000g. The aqueous phase was then used for MS. A sample containing only buffer and no protein was subjected to the same treatment for reference. The samples were then mixed with four volumes of injection buffer [90% acetonitrile, 9% methanol, 50 mM ammonium acetate (pH 7)] and directly injected using a Hamilton syringe in Synapt G2-SI HDMS (Waters) in negative mode and using the T-Wave IMS.

All density and structure representations were generated using UCSF ChimeraX (58). Difference densities were calculated in ChimeraX using the volume subtract command. Local resolutions were estimated using cryoSPARC v2 (Structura Biotechnology Inc.). The electrostatic surface representation of Rictor was generated using APBS [Adaptive Poisson-Boltzmann Solver (59)]. Multiple sequence alignment was performed using Clustal Omega (60) and visualized with Espript (61). Conservation analysis was done with AL2CO (62) and visualized in UCSF ChimeraX (58).

Acknowledgments: We thank T. Sharpe at the Biophysics facility and A. Schmidt at the Proteomics Core Facility of Biozentrum and the sciCORE scientific computing facility, all from University of Basel. We thank M. Leibundgut for advice with model building, A. Jomaa and S. Mattei for advice on cryo-EM data processing, the ETH scientific center for optical and electron microscopy (ScopeM), and, in particular, M. Peterek and P. Tittmann for technical support. We are indebted to E. Laczko and J. Hu of the Functional Genomics Center Zrich for the help with mass spectrometry. We thank I. Lui and T. Leonard (Max F. Perutz Laboratories, Vienna) for providing (Delta-PH) Akt1 protein. Funding: F.M. and K.B. are recipients of a fellowship from the Biozentrum International PhD program. This work was supported by the Swiss National Science Foundation (SNSF) via the National Center of Excellence in RNA and Disease (project funding 138262) to N.B. and M.N.H. and SNSF project funding 179323 and 177084 to T.M. Author contributions: A.S. designed the experiments, prepared the sample for cryo-EM, and carried out data processing and structure modeling. A.S. and D.B. performed data collection. F.M. designed the experiments; cloned Akt1, mTORC2 mutants, and Rictor mutants; expressed and purified proteins; and performed the activity assays and the nano-DSF measurements. E.S. established the mTORC2 purification procedure. S.I. cloned mTORC2 and contributed to data analysis and manuscript preparation. M.S. performed the in-cell analysis of mTORC2 activity. K.B. and M.S. performed the KO/KD of MINPP1 and IPPK. A.S., F.M., D.B., S.I., N.B., M.N.H., and T.M. participated in the writing of the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. The high-resolution cryo-EM map of the half-mTORC2 (density C) and full-mTORC2 (density A) has been deposited in the Electron Microscopy Data Bank as EMD-11492 and EMD-11488, respectively, while the corresponding models are in the Protein Data Bank as PDB ID 6ZWO and 6ZWM. In addition, the density of mTORC2 in the absence of ATPS (density F), as well as the densities showing extra density (densities G and H) were deposited in the Electron Microscopy Data Bank as EMD-11489, EMD-11491, and EMD-11490, respectively. Plasmid MX01 is available from Addgene. Requests for materials should be addressed to T.M.

The rest is here:
The 3.2- resolution structure of human mTORC2 - Science Advances

Are Proteins Attracted to Function? – Discovery Institute

Photo: Douglas Axe.

Doug Axe showed that functional space is a tiny fraction of sequence space in proteins. Evolutionists think they found a shortcut as simple as dropping down a funnel. Proteins dont have to search all of sequence space at random; a ring attractor pulls them down the thermodynamic funnel into functional glory land.

Richard Dawkins has been criticized for years now for his Weasel analogy (see Jonathan Witts critique). And yet the myth lives on. Miracles happen with the words, It evolves! while waving the magic wand, Natural Selection. Heres a new instance involving protein folds.

Dawkinss main error was with setting a target sequence for random letters (the Hamlet sequence Methinks it is like a weasel), and then preserving the randomly changing letters that matched the target. Natural selection as Darwin envisioned it has no target sequence. Each step must be functional, or it is not selected. All the intermediate phrases in Dawkinss computer simulation were gibberish. They had no function in language. They would never converge on the target phrase by unguided natural processes.

The same is true with random sequences of amino acids, called polypeptides. They have no function and are not called proteins or enzymes unless and until they fold into a functional shape. Before considering the following hypothesis by two chemists, remember that without guidance from genes, polypeptides fall into the vast neverland called sequence space where nothing happens (the amino acids, furthermore, must be left-handed, or homochiral). Functional space is but a tiny fraction of sequence space. Doug Axe discussed this in his book Undeniable, based on his own research at Cambridge. He experimentally determined how much change was necessary to break a functional protein with mutations. It led to his estimate that a random polypeptide 150 amino acids in length, which is modest for a protein, has only a 1 in 1074 chance of arriving at a functional fold. That probability drops to an impossible 1 in 10148 chance if the sequence must be homochiral, and even lower if the sequence also has to consist of only peptide bonds. In short, it would be a miracle.

In their paper in PNAS, Funneled energy landscape unifies principles of protein binding and evolution, Zhiqiang Yan and Jin Wang think they have found a shortcut to the miraculous. Natural selection will push the polypeptide down a thermodynamic funnel, like a golfer putting a ball into the cup. Why? Because, clearly, proteins have evolved. Anything that has evolved would have had the magic wand of natural selection to do the magic.

Most proteins have evolved to spontaneously fold into native structure and specifically bind with their partners for the purpose of fulfilling biological functions. According to Darwin, protein sequences evolve through random mutations, and only the fittest survives. The understanding of how the evolutionary selection sculpts the interaction patterns for both biomolecular folding and binding is still challenging. In this study, we incorporated the constraint of functional binding into the selection fitness based on the principle of minimal frustration for the underlying biomolecular interactions. Thermodynamic stability and kinetic accessibility were derived and quantified from a global funneled energy landscape that satisfies the requirements of both the folding into the stable structure and binding with the specific partner. The evolution proceeds via a bowl-like evolution energy landscape in the sequence space with a closed-ring attractor at the bottom. The sequence space is increasingly reduced until this ring attractor is reached. The molecular-interaction patterns responsible for folding and binding are identified from the evolved sequences, respectively. The residual positions participating in the interactions responsible for folding are highly conserved and maintain the hydrophobic core under additional evolutionary constraints of functional binding. The positions responsible for binding constitute a distributed network via coupling conservations that determine the specificity of binding with the partner. This work unifies the principles of protein binding and evolution under minimal frustration and sheds light on the evolutionary design of proteins for functions. [Emphasis added.]

Methinks these are weasel words. This is like the following syllogism. Major premise: Everything evolves by natural selection. Minor premise: Proteins occupy a tiny fraction of sequence space that permits folding and binding to specific partners. Conclusion: Natural selection pushed proteins to fulfill these constraints. Anything circular here? What if one does not accept the major premise?

To make their point, Yan and Wang know that they have to satisfy the laws of thermodynamics, which militate against functional folds by accident. Sure enough, the paper has lovely equations. But if the premise is wrong, equations only provide window dressing on a fake storefront. Here is the weasel-like target sequence:

To realize the principle of minimal frustration in protein evolution, one of the typical naturally occurring protein domains (WW domain) and its binding complex were chosen as the evolution model. WW domains preferably bind Pro-rich peptide. The native structure of the binding complex was considered as the evolved and functional structures (SI Appendix, Fig. S2). The evolution simulation is to mimic how nature selects and optimizes the sequences of the WW domain, which can spontaneously fold and preferably bind to the specific Pro-rich peptide.

Their principle of minimal frustration refers to optimization of protein sequences. The principle is useful for analyzing proteins, but not for accounting how they became optimized.

The principle of minimal frustration has been fruitful in illustrating how the global pattern of interactions determines thermodynamic stability and kinetic accessibility of protein folding and binding. The principle requires that energetic conflicts are minimized in folded native states, so that a sequence can spontaneously fold. Because of the functional necessity, naturally occurring sequences are actually in the tradeoff for coding the capacity to simultaneously satisfy stable folding and functional binding. From the view of localized frustration, naturally occurring proteins maintain a conserved network of minimally frustrated interactions at the hydrophobic core. In contrast, highly frustrated interactions tend to be clustered on the surface, often near binding sites that become less frustrated upon binding. A natural question is how the evolution sculpts the interaction patterns that conflict with the overall folding of minimal frustration but are specific for protein binding.

This principle is an ID principle: proteins are sculpted to have stable cores, but flexible surfaces. They are optimized for this. To make evolution the sculptor begs the question. Its like saying, proteins must fulfill requirements for thermodynamic stability and kinetic accessibility; therefore, evolution fulfilled these requirements. Its like saying, We take minimal frustration to be a measure of fitness, and since natural selection always moves toward higher fitness, proteins evolved the observed optimization. How do they not recognize the circular reasoning here? They are following a principle of maximal frustration for critical thinkers! Its incredible that this kind of circular argument was published in the premiere journal of the National Academy of Sciences and survived the editing scrutiny of David A. Weitz of Harvard.

Protein function is the ultimate goal of protein evolution via mutagenesis for survival. This work has proposed and quantified the selection fitness of protein evolution with the principle of minimal frustration. The selection fitness of thermodynamic stability and kinetic accessibility incorporates both folding and binding requirements. Driven by the selection fitness, the evolution dynamics in sequence space can be depicted and visualized as a bowl-like energy landscape where the sequence space is increasingly reduced until the closed-ring attractor is reached at the bottom. The evolved sequences located in the basin of the attractor faithfully reproduce the interaction patterns as those extracted from naturally occurring sequences. The consistency validates the principle of minimal frustration as the selection fitness of protein evolution. To fulfill the folding and function, evolution sculpts the interaction patterns with the minimal-frustration principle to develop the hydrophobic core for folding and the coupling network for functional binding.

Comparing this to Dawkins Weaselology, this is like saying, The goal of sentences is to express meaning. Driven by this selection fitness, evolution dynamics guarantee that random letters will fall through a bowl-like semantics landscape where the randomness is reduced until a closed ring of meaningful sentences naturally occurs. The fact that natural sentences convey meaning validates this principle. Evolution sculpts meaning from random letters because it must, and lo and behold, it does. Aaagggh! How does this notion pass peer review?

To make their circularity seem practical, they show what else could be done by reasoning in a circle in the wide-angle view:

In addition, the evolution of a protein binding/assembling system generally involves the evolution of each binding/assembling partner. Therefore, the evolution of one partner is constrained or coupled to the evolution of its partners, i.e., coevolution of the partners, such as a toxinantitoxin system. In this case, the selection fitness of protein evolution involves the constraints not only from its own folding and binding but also from those of its partners. The study of this more complex issue would bridge the evolution of a single protein to the evolution of a protein network.

The whole world is circular. Isnt that a useful idea!

This kind of reasoning is not limited to this paper. Five authors, including Joseph W. Thornton (whom Michael Behe says threw a monkey wrench into Darwinian evolution), wrote a preprint on bioRxiv with similar fallacies. In Chance, contingency, and necessity in the experimental evolution of ancestral proteins, they assert that varieties of BCL-2 (an anti-apoptosis protein) arrived at their optimum fitness by convergent evolution, even though they recognize that there was no way to expect that because of the inevitability of chance and contingency in unguided natural processes:

Finally, our observations suggest that the sequence-structure-function associations apparent in sequence alignments are, to a significant degree, the result of shared but contingent constraints that were produced by chance events during history. Present-day proteins are physical anecdotes of a particular history: they reflect the interaction of accumulated chance events during descent from common ancestors with necessity imposed by physics, chemistry and natural selection. Apparent design principles in extant or evolved proteins express not how things must be or even how they would be best but rather the contingent legacy of the constraints and opportunities that those molecules just happen to have inherited.

Once again, they treat natural selection as a sculptor with a guiding hand. The constraints to get sequences that work must have chosen working products out of the vast sea of possibilities. They evolved because they evolved. They look designed, but the design principles are only apparent.

Dawkins would be pleased that his fallacy continues to be fruitful. His critics worry about the overpopulation of weasels in science. Hawks, flying overhead the infested area, casting a wide view over the creatures running in circles, make good weasel predators.

Link:
Are Proteins Attracted to Function? - Discovery Institute

Cross-reactive neutralization of SARS-CoV-2 by serum antibodies from recovered SARS patients and immunized animals – Science Advances

Abstract

The current coronavirus disease 2019 (COVID-19) pandemic is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), a novel coronavirus genetically close to SARS-CoV. To investigate the effects of previous SARS-CoV infection on the ability to recognize and neutralize SARS-CoV-2, we analyzed 20 convalescent serum samples collected from individuals infected with SARS-CoV during the 2003 SARS outbreak. All patient sera reacted strongly with the S1 subunit and receptor binding domain (RBD) of SARS-CoV; cross-reacted with the S ectodomain, S1, RBD, and S2 proteins of SARS-CoV-2; and neutralized both SARS-CoV and SARS-CoV-2 S proteindriven infections. Analysis of antisera from mice and rabbits immunized with a full-length S and RBD immunogens of SARS-CoV verified cross-reactive neutralization against SARS-CoV-2. A SARS-CoVderived RBD from palm civets elicited more potent cross-neutralizing responses in immunized animals than the RBD from a human SARS-CoV strain, informing strategies for development of universal vaccines against emerging coronaviruses.

The global outbreak of the coronavirus disease 2019 (COVID-19) was caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which is a new coronavirus (CoV) genetically close to SARS-CoV that emerged in 2002 (13). As of 25 May 2020, a total of 5,307,298 confirmed COVID-19 cases, including 342,070 deaths, have been reported from 216 countries or regions, and the numbers are still growing rapidly (https://who.int). Unfortunately, even though 17 years passed, we have not developed effective prophylactics and therapeutics in preparedness for the reemergence of SARS or SARS-like CoVs. A vaccine is urgently needed to prevent the human-to-human transmission of SARS-CoV-2.

Like SARS-CoV and many other CoVs, SARS-CoV-2 uses its surface spike (S) glycoprotein to gain entry into host cells (46). Typically, the S protein forms a homotrimer with each protomer consisting of S1 and S2 subunits. The N-terminal S1 subunit is responsible for virus binding to the cellular receptor angiotensin-converting enzyme 2 (ACE2) through an internal receptor binding domain (RBD) that is capable of functional folding independently, whereas the membrane-proximal S2 subunit mediates membrane fusion events. While SARS-CoV-2 and SARS-CoV share about 80% homology in full-length genome sequences, their S proteins have about 76% amino acid identity (2, 3). The RBD sequences of the two viruses are only about 74% identical, with most mutations occurring in the receptor-binding motifs (RBMs) (~50% amino acid identity). It was found that the ACE2-binding affinity of the SARS-CoV-2 RBD is 10- to 20-fold higher than that of the SARS-CoV RBD, which may contribute to the higher transmissibility of SARS-CoV-2 (7). Very recently, the prefusion structure of the SARS-CoV-2 S protein was determined by cryoelectron microscopy, which revealed an overall similarity to that of SARS-CoV (5, 7); the crystal structure of the SARS-CoV-2 RBD in complex with ACE2 was also determined by several independent groups, and the residues or motifs critical for the higher-affinity RBD-ACE2 interaction were identified (810). As seen, the SARS-CoV-2 RBD binds ACE2 in the same orientation with the SARS-CoV RBD and relies on conserved, mostly aromatic, residues. The structures have also provided evidence to support a mechanism of infection triggering that is thought to be conserved among the Coronaviridae, wherein the S protein undergoes distinct conformational states with the RBD closed (receptor-inaccessible) or opened (receptor-accessible).

The S protein of CoVs is also a main target of neutralizing antibodies (nAbs), thus being considered an immunogen for vaccine development (5, 11). During the SARS-CoV outbreak in 2002, we took immediate actions to characterize the immune responses in infected SARS patients and in inactivated virus vaccine- or S proteinimmunized animals (1220). We demonstrated that the S protein RBD dominates the nAb response against SARS-CoV infection and thus proposed an RBD-based vaccine strategy (11, 1522). Our follow-up studies verified a potent and persistent anti-RBD response in recovered SARS patients (2325). Although SARS-CoV-2 and SARS-CoV share substantial genetic and functional similarities, their S proteins, especially in the RBD sequences, display relatively larger divergences. Toward developing vaccines and immunotherapeutics against emerging CoVs, it is fundamentally important to characterize the antigenic cross-reactivity between SARS-CoV-2 and SARS-CoV.

A panel of serum samples collected from 20 patients who recovered from SARS-CoV infection was analyzed for the antigenic cross-reactivity with SARS-CoV-2. First, we examined the convalescent sera by a commercial diagnostic enzyme-linked immunosorbent assay (ELISA) kit, which uses a recombinant nucleocapsid (N) protein of SARS-CoV-2 as detection antigen. As shown in Fig. 1A, all the serum samples at a 1:100 dilution displayed high reactivity, verifying that the N antigen is highly conserved between SARS-CoV and SARS-CoV-2. As tested by ELISA, each of the patient sera also reacted with the SARS-CoV S1 subunit and its RBD strongly (Fig. 1B). Then, we determined the cross-reactivity of the patient sera with four recombinant protein antigens derived from the S protein of SARS-CoV-2, including S ectodomain (designated S), S1 subunit, RBD, and S2 subunit. As shown in Fig. 1C, all the serum samples also reacted strongly with the S and S2 proteins, but they were less reactive with the S1 and RBD proteins.

(A) Reactivity of sera from 20 recovered patients with SARS-CoV (P01 to P20) with the nucleoprotein (N) of SARS-CoV-2 was measured by a commercial ELISA kit. The positive (pos) or negative (neg) control serum sample provided in the kit was collected from a convalescent SARS-CoV-2infected individual or healthy donor. (B) Reactivity of convalescent SARS sera with the recombinant S1 and RBD proteins of SARS-CoV. (C) Reactivity of convalescent SARS sera with the S ectodomain (designated S), S1, RBD, and S2 proteins of SARS-CoV-2. Serum samples from two healthy donors were used as negative control (Ctrl-1 and Ctrl-2). The experiments were performed with duplicate samples and repeated three times, and data are shown as means with SDs. OD450, optical density at 450 nm.

Limited by facility that can handle authentic viruses, we developed a pseudovirus-based single-cycle infection assay to determine the cross-neutralizing activity of the convalescent SARS sera on SARS-CoV and SARS-CoV-2. A control lentivirus was pseudotyped with vesicular stomatitis virus G protein (VSV-G). Initially, the serum samples were analyzed at a 1:20 dilution. As shown in Fig. 2A, all the sera efficiently neutralized both the SARS-CoV and SARS-CoV-2 pseudoviruses to infect 293T/ACE2 cells, and in comparison, each serum had lower efficiency in inhibiting SARS-CoV-2 as compared to SARS-CoV. None of the immune sera showed appreciable neutralizing activity on VSV-G pseudovirus. The neutralizing titer for each patient serum was then determined. As shown in Fig. 2B, the patient sera could neutralize SARS-CoV with titers ranging from 1:120 to 1:3240 and could cross-neutralized SARS-CoV-2 with titers ranging from 1:20 to 1:360. In a highlight, the patient P08 serum had the highest titer to neutralize SARS-CoV (1:3240) when it neutralized SARS-CoV-2 with a titer of 1:120; the patient P13 serum showed the highest titer on SARS-CoV-2 (1:360) when it had a 1:1080 titer to efficiently neutralize SARS-CoV.

(A) Neutralizing activities of convalescent patient sera (1:20 dilution) against SARS-CoV, SARS-CoV-2, and VSV-G control were tested by a single-cycle infection assay. (B) Neutralizing titers of each of the convalescent patient sera on the three pseudotypes were measured. The experiments were performed with triplicate samples and repeated three times, and data are shown as means with SDs.

To comprehensively characterize the cross-reactivity between the S proteins of SARS-CoV and SARS-CoV-2, we generated mouse antisera against the S protein of SARS-CoV by immunization. Here, three mice (M-1, M-2, and M-3) were immunized with a recombinant full-length S protein in the presence of MPL-TDM adjuvant (monophosphoryl lipid A plus trehalose dicorynomycolate), while two mice (M-4 and M-5) were immunized with the S protein plus alum adjuvant (fig. S1). Binding of antisera to diverse S antigens were initially examined by ELISA. As shown in Fig. 3A, the mice immunized by the S protein with the MPL-TDM adjuvant developed relatively higher titers of antibody responses as compared to the two mice with the alum adjuvant. It was expected that the adjuvanticity of alum formulation was weaker than that of MPL-TDM. Apparently, each of the mouse antisera had high cross-reactivity with the SARS-CoV-2 S and S2 proteins, but the cross-reactive antibodies specific for the SARS-CoV-2 S1 and RBD were relatively lower except that in mouse M-3. Subsequently, the neutralizing capacity of mouse anti-S sera was measured with pseudoviruses. As shown in Fig. 3 (B to F), all the antisera, diluted at 1:40, 1:160, or 1:640, potently neutralized SARS-CoV, and consistently, they were able to cross-neutralize SARS-CoV-2 although with reduced capacity relative to SARS-CoV.

(A) Binding activity of mouse anti-S sera at a 1:100 dilution to SARS-CoV (S1 and RBD) and SARS-CoV-2 (S, S1, RBD, and S2) antigens was determined by ELISA. A healthy mouse serum was tested as control. (B to F) Neutralizing activity of mouse anti-S sera at indicated dilutions against SARS-CoV, SARS-CoV-2, and VSV-G pseudoviruses was determined by a single-cycle infection assay. The experiments were performed in triplicate and repeated three times, and data are shown as means with SDs. Statistical significance was tested by two-way ANOVA with Dunnett posttest. **P 0.01 and ***P 0.001.

As the S protein RBD dominates the nAb response to SARS-CoV, we sought to characterize the RBD-mediated cross-reactivity and neutralization on SARS-CoV-2. To this end, we first generated mouse anti-RBD sera by immunization with two RBD-Fc fusion proteins: one encoding the RBD sequence of a palm civet SARS-CoV strain SZ16 (SZ16-RBD) and the second one with the RBD sequence of a human SARS-CoV strain GD03 (GD03-RBD). Both the fusion proteins were expressed in 293T cells and purified to apparent homogenicity (fig. S1). As shown in Fig. 4A, all eight mice developed robust antibody responses against the SARS-CoV S1 and RBD, and in comparison, four mice (M-1 to M-4) immunized with SZ16-RBD exhibited higher titers of antibody responses than the mice (M-5 to M-8) immunized with GD03-RBD. Each of the anti-RBD sera cross-reacted well with the S protein of SARS-CoV-2, suggesting that SARS-CoV and SARS-CoV-2 do share antigenically conserved epitopes in the RBD sites. Noticeably, while the SZ16-RBD immune sera also reacted with the SARS-CoV-2 S1 and RBD antigens, the cross-reactivity of the GD03-RBD immune sera was low. However, while the mouse anti-RBD sera at 1:50 dilutions were measured with increased coating antigens in ELISA, they reacted with the SARS-CoV-2 S1 and RBD efficiently, which verified the cross-reactivity (Fig. 4B). Similarly, the neutralizing activity of mouse antisera was determined by pseudovirus-based single-cycle infection assay. As shown in Fig. 4 (C and D), both the SZ16-RBD and GD03-RBDspecific antisera displayed very potent activities to neutralize SARS-CoV; they also cross-neutralized SARS-CoV-2 with relatively lower efficiencies. As judged by the neutralizing activity at the highest serum dilution, the SZ16-RBD antisera were more potent than the GD03-RBD antisera in neutralizing SARS-CoV; however, the two antisera had no significant difference in neutralizing SARS-CoV-2 (Fig. 4, E and F).

(A) Binding activity of mouse antisera at a 1:100 dilution to SARS-CoV (S1 and RBD) and SARS-CoV-2 (S, S1, and RBD) antigens was determined by ELISA. A healthy mouse serum was tested as control. (B) The cross-reactivity of mouse antisera with the SARS-CoV-2 S1 and RBD proteins. The antisera were diluted at 1:50, and the S1 and RBD antigens were coated at 100 ng per ELISA plate well. (C and D) Neutralizing activities of mouse antisera at indicated dilutions against SARS-CoV, SARS-CoV-2, and VSV-G pseudoviruses were determined by a single-cycle infection assay. The experiments were performed in triplicate and repeated three times, and data are shown as means with SDs. (E and F) Comparison of neutralizing activities of the mouse antiSZ16-RBD and antiGD03-RBD sera. Statistical significance was tested by two-way ANOVA with Dunnett posttest. ns, not significant. *P 0.05, **P 0.01, and ***P 0.001.

We further developed rabbit antisera by immunizations, in which two rabbits were immunized with SZ16-RBD (R-1 and R-2) or with GD03-RBD (R-3 and R-4). Each RBD protein elicited antibodies highly reactive with both the SARS-CoV and SARS-CoV-2 antigens (Fig. 5A), which were different from their immunizations in mice. As expected, all of the rabbit antisera potently neutralized SARS-CoV and SARS-CoV-2 in a similar profile with that of the mouse anti-S and anti-RBD sera (Fig. 5, B and C). Obviously, the neutralizing activity of rabbit antiSZ16-RBD sera against both the viruses was higher than that of the rabbit antiGD03-RBD sera (Fig. 5, D and E). Together, the results verified that the SARS-CoV S protein and its RBD immunogens can induce cross-neutralizing antibodies toward SARS-CoV-2 by vaccination.

(A) Binding activity of rabbit antisera at a 1:100 dilution to SARS-CoV (S1 and RBD) and SARS-CoV-2 (S protein and RBD) antigens was determined by ELISA. A healthy rabbit serum was tested as control. (B and C) Neutralizing activities of rabbit antisera or control serum at indicated dilutions on SARS-CoV, SARS-CoV-2, and VSV-G pseudoviruses were determined by a single-cycle infection assay. The experiments were done in triplicate and repeated three times, and data are shown as means with SDs. (D and E) Comparison of neutralizing activities of the rabbit antiSZ16-RBD and antiGD03-RBD sera. Statistical significance was tested by two-way ANOVA with Dunnett posttest. *P 0.05, **P 0.01, and ***P 0.001.

To validate the observed cross-reactive neutralization and explore the underlying mechanism, we purified anti-RBD antibodies from the rabbit antisera above. As shown in Fig. 6 (A and B), both purified rabbit antiSZ16-RBD and antiGD03-RBD antibodies reacted strongly with the SARS-CoV RBD protein and cross-reacted with the SARS-CoV-2 S and RBD but not S2 proteins in a dose-dependent manner. Moreover, the purified antibodies dose-dependently neutralized SARS-CoV and SARS-CoV-2 but not VSV-G (Fig. 6, C and D). Consistent with their antisera, the rabbit antiSZ16-RBD antibodies were more active than the rabbit antiGD03-RBD antibodies against both SARS-CoV and SARS-CoV-2 (Fig. 6, E and F). Next, we investigated whether the rabbit anti-RBD antibodies block RBD binding to 293T/ACE2 cells by flow cytometry. As expected, both the SARS-CoV and SARS-CoV-2 RBD proteins could bind to 293T/ACE2 cells in a dose-dependent manner and, in line with previous findings, that the RBD of SARS-CoV-2 bound to ACE2 more efficiently (fig. S2). Unexpectedly, the antibodies purified from SZ16-RBDimmunized rabbits (R-1 and R-2) potently blocked the binding of both the RBD proteins, whereas the antibodies from GD03-RBDimmunized rabbits (R-3 and R-4) had no such blocking functionality except a high concentration of the rabbit R-3 antibody on the SARS-CoV RBD binding (Fig. 7).

Binding titers of purified rabbit antiSZ16-RBD (A) and antiGD03-RBD (B) antibodies (Abs) to the SARS-CoV (RBD) and SARS-CoV-2 (S, RBD, and S2) antigens were determined by ELISA. A healthy rabbit serum was tested as control. (C and D) Neutralizing titers of purified rabbit antiSZ16-RBD and antiGD03-RBD antibodies on SARS-CoV, SARS-CoV-2, and VSV-G pseudoviruses were determined by a single-cycle infection assay. The experiments were done in triplicate and repeated three times, and data are shown as means with SDs. (E and F) Comparison of neutralizing activities of the rabbit antiSZ16-RBD and antiGD03-RBD antibodies.

(A) Blocking activity of rabbit anti-RBD antibodies on the binding of SARS-CoV RBD (first two panels) or SARS-CoV-2 RBD (last two panels) to 293T/ACE2 cells was determined by flow cytometry. FITC-A, fluorescein isothiocyanate-labeled concanavalin A. (B) Purified rabbit anti-RBD antibodies inhibited the RBD-ACE2 binding dose-dependently. The experiments were repeated three times, and data are shown as means with SDs. Statistical significance was tested by two-way ANOVA with Dunnett posttest. *P 0.05 and **P 0.01.

To develop effective vaccines and immunotherapeutics against emerging CoVs, the antigenic cross-reactivity between SARS-CoV-2 and SARS-CoV is a key scientific question that needs to be addressed as soon as possible. However, after the SARS-CoV outbreak more than 17 years ago, there are very limited blood samples from SARS-CoVinfected patients available for such studies. At the moment, Hoffmann et al. (26) analyzed three convalescent patient with SARS sera and found that both SARS-CoV-2 and SARS-CoV S protein-driven infections were inhibited by diluted sera, but the inhibition of SARS-CoV-2 was less efficient; Ou et al. (27) detected one patient with SARS serum that was collected at 2 years after recovery, which showed a serum neutralizing titer of >1:80 dilution for SARS-CoV pseudovirus and of 1:40 dilution for SARS-CoV-2 pseudovirus. While these studies supported the cross-neutralizing activity of the convalescent SARS sera on SARS-CoV-2, a just published study with the plasma from seven SARS-CoVinfected patients suggested that cross-reactive antibody binding responses to the SARS-CoV-2 S protein did exist, but cross-neutralizing responses could not be detected (28). In this study, we first investigated the cross-reactivity and neutralization with a panel of precious immune sera collected from 20 recovered SARS patients. As shown, all the patient sera displayed high titers of antibodies against the S1 and RBD proteins of SARS-CoV and cross-reacted strongly with the S protein of SARS-CoV-2. In comparison, the patient sera had higher reactivity with the S2 subunit of SARS-CoV-2 relative to its S1 subunit and RBD protein, consistent with a higher sequence conservation between the S2 subunits of SARS-CoV-2 and SARS-CoV than that of their S1 subunits and RBDs (3, 5). Each of the patient sera could cross-neutralize SARS-CoV-2 with serum titers ranging from 1:20 to 1:360 dilutions, verifying the cross-reactive neutralizing activity of the patient with SARS sera on the S protein of SARS-CoV-2.

Now, two strategies are being explored for developing vaccines against emerging CoVs. The first one is based on a full-length S protein or its ectodomain, while the second uses a minimal but functional RBD protein as vaccine immunogen. Our previous studies revealed that the RBD site contains multiple groups of conformation-dependent neutralizing epitopes: Some epitopes are critically involved in RBD binding to the cell receptor ACE2, whereas other epitopes have a neutralizing function but do not interfere with the RBD-ACE2 interaction (15, 18). Most neutralizing monoclonal antibodies (mAbs) previously developed against SARS-CoV target the RBD epitopes, while a few are directed against the S2 subunit or the S1/S2 cleavage site (29, 30). The cross-reactivity of such mAbs with SARS-CoV-2 has been characterized, and it was found that many SARS-CoVneutralizing mAbs exhibit no cross-neutralizing capacity (8, 31). For example, CR3022, a nAb isolated from a convalescent patient with SARS, cross-reacted with the RBD of SARS-CoV-2 but did not neutralize the virus (31, 32). Nonetheless, a new human anti-RBD mAb, 47D11, has just been isolated from transgenic mice immunized with a SARS-CoV S protein, which neutralizes both SARS-CoV-2 and SARS-CoV (33). The results of polyclonal antisera from immunized animals are quite inconsistent. For example, Walls et al. (5) reported that plasma from four mice immunized with a SARS-CoV S protein could completely inhibit SARS-CoV pseudovirus and reduced SARS-CoV-2 pseudovirus to ~10% of control, thus proposing that immunity against one virus of the sarbecovirus subgenus can potentially provide protection against related viruses; two rabbit antisera raised against the S1 subunit of SARS-CoV also reduced SARS-CoV-2 Sdriven cell entry although with lower efficiency compared to SARS-CoV S (26). Moreover, four mouse antisera against the SARS-CoV RBD cross-reacted efficiently with the SARS-CoV-2 RBD and neutralized SARS-CoV-2, suggesting the potential to develop a SARS-CoV RBDbased vaccine preventing SARS-CoV-2 (34). Differently, it was reported that plasma from mice infected or immunized by SARS-CoV failed to neutralize SARS-CoV-2 infection in Vero E6 cells (28), and mouse antisera raised against the SARS-CoV RBD were even unable to bind to the S protein of SARS-CoV-2 (8). In the present studies, several panels of antisera against the SARS-CoV S and RBD proteins were comprehensively characterized. Although the use of pseudovirus-based neutralization assay might not fully reflect the complexity of authentic SARS-CoV-2 infection, our results, altogether, did provide reliable data to validate the cross-reactivity and cross-neutralization between SARS-CoV and SARS-CoV-2. Meaningfully, this work found that the RBD proteins derived from different SARS-CoV strains can elicit antibodies with unique functionalities: While the RBD from a palm civet SARS-CoV (SZ16) induced potent antibodies capable of blocking the RBD-receptor binding, the antibodies elicited by the RBD derived from a human strain (GD03) had no such effect despite their neutralizing activities. SZ16-RBD shares an overall 74% amino acid sequence identity with the RBD of SARS-CoV-2, when their internal RBMs display more marked substitutions (~50% sequence identity); however, SZ16-RBD and GD03-RBD only differ from three amino acids, all located within the RBM (fig. S3). Further research is needed to determine how these mutations change the antigenicity and immunogenicity of the S protein and RBD immunogens.

Three more questions invite further investigation. First, it would be intriguing to know whether individuals who recovered from previous SARS-CoV infection can direct their acquired SARS-CoV immunity against SARS-CoV-2 infection. To address this question, an epidemiological investigation of populations exposed to SARS-CoV-2 would provide valuable insights. Second, it would be important to determine whether a universal vaccine can be rationally designed by engineering the S protein RBD sequences. Third, although antibody-dependent infection enhancement was not observed during our studies with the human and animal serum antibodies, the possibility of such effects should be carefully addressed in vaccine development.

Two RBD-Fc fusion proteins, which contain the RBD sequence of Himalayan palm civet SARS-CoV strain SZ16 (accession number: AY304488.1) or the RBD sequence of human SARS-CoV strain GD03T0013 (AY525636.1, denoted GD03) linked to the Fc domain of human immunoglobulin G1 (IgG1), were expressed in transfected 293T cells and purified with protein ASepharose 4 Fast Flow in our laboratory as previously described (15). A full-length S protein of SARS-CoV Urbani (AY278741) was expressed in expressSF+ insect cells with recombinant baculovirus D3252 by the Protein Sciences Corporation (Bridgeport, CT, USA) (16). A panel of recombinant proteins with a C-terminal polyhistidine (His) tag, including S1 and RBD of SARS-CoV (AAX16192.1) and S ectodomain (S-ecto), S1, RBD, and S2 of SARS-CoV-2 (YP_009724390.1), were purchased from the Sino Biological Company (Beijing, China) and characterized for quality control by SDSpolyacrylamide gel electrophoresis (fig. S4).

Twenty patients with SARS were enrolled in March 2003 for a follow-up study at the Peking Union Medical College Hospital, Beijing. Serum samples were collected from recovered patients at 3 to 6 months after discharge, with the patients written consent and the approval of the ethics review committee (23, 24). The samples were stored in aliquots at 80C and were heat-inactivated at 56C before performing experiments.

Multiple immunization protocols were conducted in compliance with the Institutional Animal Care and Use Committee guidelines and are summarized in fig. S1B. First, five Balb/c mice (6 weeks old) were subcutaneously immunized with 20 g of full-length S protein resuspended in phosphate-buffered saline (PBS; pH 7.2) in the presence of MPL-TDM adjuvant or alum adjuvant (Sigma-Aldrich). Second, eight Balb/c mice (6 weeks old) were subcutaneously immunized with 20 g of SZ16-RBD or GD03-RBD fusion proteins and MPL-TDM adjuvant. The mice were boosted two times with 10 g of the same antigens and the MPL-TDM adjuvants at 3-week intervals. Third, four New Zealand White rabbits (12 weeks old) were immunized intradermally with 150 g of SZ16-RBD or GD03-RBD resuspended in PBS (pH 7.2) in the presence of Freunds complete adjuvant and boosted two times with 150 g of the same antigens and incomplete Freunds adjuvant at 3-week intervals. Mouse and rabbit antisera were collected and stored at 40C.

Binding activity of serum antibodies with diverse S protein antigens was detected by ELISA. In brief, 50 or 100 ng of a purified recombinant protein (SARS-CoV S1 or RBD and SARS-CoV-2 S-ecto, S1, RBD, or S2) was coated into a 96-well ELISA plate overnight at 4C. Wells were blocked with 5% bovine serum albumin in PBS for 1 hour at 37C, followed by incubation with diluted antisera or purified rabbit antibodies for 1 hour at 37C. A diluted horseradish peroxidaseconjugated goat anti-human, mouse, or rabbit IgG antibody was added for 1 hour at room temperature. Wells were washed five times between each step with 0.1% Tween 20 in PBS. Wells were developed using 3,3,5,5-tetramethylbenzidine and read at 450 nm after termination with 2 M H2SO4.

Neutralizing activity of serum antibodies was measured by pseudovirus-based single-cycle infection assay as previously described (35). The pseudovirus particles were prepared by cotransfecting 293T cells with a backbone plasmid (pNL4-3.luc.RE) that encodes an Env-defective, luciferase reporter-expressing HIV-1 genome and a plasmid expressing the S protein of SARS-CoV-2 (IPBCAMS-WH-01; accession number: QHU36824.1) or SARS-CoV (GD03T0013) or the VSV-G. Cell culture supernatants containing virions were harvested 48 hours after transfection, filtrated, and stored at 80C. To measure the neutralizing activity of serum antibodies, a pseudovirus was mixed with an equal volume of serially diluted sera or purified antibodies and incubated at 37C for 30 min. The mixture was then added to 293T/ACE2 cells at a density of 104 cells/100 l per plate well. After culture at 37C for 48 hours, the cells were harvested and lysed in reporter lysis buffer, and luciferase activity (relative luminescence unit) was measured using luciferase assay reagents and a luminescence counter (Promega, Madison, WI). Percent inhibition of serum antibodies compared to the level of the virus control subtracted from that of the cell control was calculated. The highest dilution of the serum sample that reduced infection by 50% or more was considered to be positive.

Blocking activity of purified rabbit anti-RBD antibodies on the binding of RBD proteins with a His tag to 293T/ACE2 cells was detected by flow cytometry assay. Briefly, SARS-CoV-2 RBD protein (2 g/ml) or SARS-CoV RBD protein (10 g/ml) was added to 4 105 cells and incubated for 30 min at room temperature. After washing two times with PBS, cells were incubated with a 1:500 dilution of Alexa Fluor 488labeled rabbit antiHis tag antibody (Cell Signaling Technology, Danvers, MA) for 30 min at room temperature. After two washes, cells were resuspended in PBS and analyzed by FACSCantoII instrument (Becton Dickinson, Mountain View, CA).

Statistical analyses were carried out using GraphPad Prism 7 Software. One-way or two-way analysis of variance (ANOVA) was used to test for statistical significance. Only P values of 0.05 or lower were considered statistically significant [P > 0.05 (ns, not significant), *P 0.05, **P 0.01, and ***P 0.001].

Acknowledgments: Funding: This work was supported by grants from the National Natural Science Foundation of China (81630061 and 82041006) and the CAMS Innovation Fund for Medical Sciences (2017-I2M-1-014). Author contributions: Conceptualization: Y. He and T.L. Formal analysis: Y.Z., D.Y., and Y. He. Investigation: Y.Z., D.Y., Y. Han, H.Y., H.C., and L.R. Resources: H.C., L.R., J.W., T.L., and Y. He. Writingoriginal draft: Y. He. Writingreview and editing: all authors. Funding acquisition: Y. He and T.L. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Read this article:
Cross-reactive neutralization of SARS-CoV-2 by serum antibodies from recovered SARS patients and immunized animals - Science Advances

Brightseed’s First Major Phytonutrient Discovery Finds Black Pepper May Help with Fatty Liver – The Spoon

Brightseed, which uses artificial intelligence (AI) to uncover previously hidden phytonutrients in plants, today announced preclinical data from its first major discovery targeting liver and metabolic health.

The discovery was made with Forager, Brightseeds AI platform that looks at plants on a molecular level to identify novel phytonutrient compounds (for example, antioxidants in blueberries). Once found, Forager then catalogs these compounds and uses that information to predict the health benefits of those compounds.

With todays announcement, Brightseeds Forager has identified phytonutrients that can help with fat accumulation in the pancreas and liver, a condition linked to obesity. Brightseed explained its findings in a press release, writing:

Using a computational approach with data from Brightseeds plant compound library, Forager identified two natural compounds with promising bioactive function, N-trans caffeoyltyramine (NTC) and N-trans-feruloyltyramine (NTF). Researchers determined that these compounds acted through a novel biological mechanism governing the accumulation and clearance of liver fat. The preclinical data was presented in the fall of 2020 as a poster session at The Liver Meeting Digital Experience hosted by American Association for the Study of Liver Diseases, and published as abstract #1679 in Hepatology: Vol 72, No S1.

The release continued:

IIn preclinical studies, NTC and NTF acted as potent HNF4a activators, promoting fat clearance from the steatotic livers of mice fed a high fat diet, by inducing lipophagy. HNF4a is a central metabolic regulator that is impaired by elevated levels of fat in the bloodstream resulting from chronic overeating. Administered in proper doses, NTC and NTF restored proper function of this central metabolic regulator, including maintaining healthy lipid and sugar levels in the bloodstream to normalize organ function. Their activities were confirmed using a cell-based human insulin promoter activation assay. Forager found NTC and NTF in over 80 common edible plant sources.

One of those plant sources, Brightseed Co-Founder and CEO, Jim Flatt told me by phone this week, is black pepper. Now, before you run out and grab your pepper grinder, there is still a lot of work that remains before the results of this discovery bear out.

First, the compounds still need to go through clinical trials to validate Brightseeds initial findings. This includes not only confirming any health benefits, but also determining the doses and best methods for administering the compounds. Then the best plant source for those compounds needs to be determined as well as the best method for compound extraction. Flatt told me that if all goes well, you can expect to see some form of supplement on the market by the end of 2022.

Even though that is a ways off, part of the reason to be excited by todays announcement is because of how little time it took Brightseed to make this particular discovery. Through its computational processes, Flatt told me his company was able to shrink what used to take years down to months. Fifteen to 20 percent of time that is computational saves us 80 percent of the time in the lab, Flatt said.

Brightseed has already analyzed roughly 700,000 compounds in the plant world for health properties and says its on track to surpass 10 million by 2025. Doing so could help unlock a number of previously unknown treatments for a number of ailments and conditions as well as general improvement to our metabolic and immuno health.

In addition to independent research such as todays findings, Brightseed also partners with major CPG brands to help them identify new applications for their products. For instance, Danone is using Brightseeds technology to help find new health benefits of soy.

Brightseeds announcement today also reinforces the bigger role AI will play in our food system. AI and machine learning is being used to do everything from turning data into cheese, to solving complex issues around protein folding.

As more discoveries using AI are made, more investment will be poured into the space, which will accelerate even more discoveries.

Related

Read more from the original source:
Brightseed's First Major Phytonutrient Discovery Finds Black Pepper May Help with Fatty Liver - The Spoon

Bootcamp allows Oswego researchers to explore fighting COVID-19 – NNY360

OSWEGO - SUNY Oswego students, faculty and recent alumni were part of a research team that spans many institutions and disciplines to research the COVID-19 pandemic resulting from the SARS-CoV-2 virus.

SUNY Oswego students Emily Fingar, Michael Kirsch and Charlotte Labrie-Cleary and recent graduates Ali Khan and Santiago Soto joined Julia Koeppe of the chemistry faculty for the weeklong bootcamp hosted by the Institute for Quantitative Biomedicine at Rutgers University this summer.

Other institutions in Oswegos group included Rochester Institute of Technology, Ursinus College, Hope College, Grand View University and Xavier University. The goal was to bring together teams of interdisciplinary researchers with complementary skills and interests to investigate the virus. Carried out completely remotely, participants interacted with experts and learned how to use various bioinformatics tools to answer pertinent research questions.

Research focused on the SARS-CoV-2 main protease (an enzyme that breaks down proteins into smaller units), which is essential for viral activity and a promising drug target. By understanding the differences in this protease resulting from the rapid evolution, researchers can move closer to developing an antiviral medicine to help COVID-19 patients, Koeppe said.

Students learned to work remotely (in Zoom and Zoom breakout rooms) with a group of their peers and a faculty mentor to study the structure and function of the main protease from the virus, Koeppe said. Students learned about computer programs used to view macromolecules such as proteins and enzymes; key principles of bioinformatics, such as sequence alignments that can show the evolution of proteins; and computer programs that model protein folding to determine three-dimensional structures.

At the end of the boot camp, all of the students gave a short presentation with their group members on some specific questions that they explored when looking at changes in the amino acid sequence of the SARS-CoV-2 protease and how they expected these changes would or would not affect the function of the protease, Koeppe said.

Preparing young researchers

Khan, a May graduate who is starting Ph.D. work in cancer biology at the University of Iowas Carver Medical School, worked in a team with two other students and Koeppe.

We were given daily tasks in which we used various structural visualizing tools to understand different mutations of coronavirus with respect to bond length, change in heat energy, etc., Khan said. There were a couple of mutations assigned per group and we had to analyze those and came up with a conclusion. We then gave a mini-presentation at the end of the week for our group about our findings.

Khan said the knowledge and interactions all were fruitful for his future plans.

This Bootcamp taught me how to interface with a scientist in a different field, Khan said. I also got an opportunity to attend various lectures which taught me the importance of research and how impactful research can be. I was also able to learn how to use visualization softwares and python programming language which will definitely come in handy in my Ph.D.

For Santiago Soto, who earned his biology degree in May and is already working in the field professionally as a clinical laboratory technologist with Acutis Diagnostics, the bootcamp helped with his important everyday work with live SARS-CoV-2 samples.

I really enjoyed the opportunity to observe the mutation and evolution process of SARS-CoV-2s over the past six months and its main protease Nsp5 while comparing it to the original viral isolate to 161 unique sequence/structure variants, Soto said. This was done by analyzing amino acid sequences using 3D atomic level structures using several bioinformatic tools. The research found Nsp5 could be a promising drug target for vaccine development, he added.

This bootcamp allowed me to better understand the use of bioinformatics/biostatistics, Soto noted. Its the base principle on being able to make identifications on the genetic basis of diseases, their desirable properties and unique adaptations. I would like to pursue sometime in the future a graduate degree and career in epidemiology, biomedical engineering or genetics, where the use of bioinformatics is constantly being used to assist in progression.

A member of Koeppes research team, senior biology and health science major Emily Fingar was immediately interested when Koeppe reached out with the opportunity. She learned Foldit, PyRosetta, and Mol visualization software programs so that we could take our assigned mutants, where we had the DNA sequence but not necessarily a structure, and force those mutations into the known protease structure, she said.

My team specifically was assigned 11 mutations in the SARS-CoV-2 main protease to characterize, Fingar said. Our goal was to model, using these programs, how each of the assigned mutations of the SARS-CoV-2 main protease might be changing at the protein level as well as the stability of that protein. We also used this data to examine if there are regions in the protein structure that are mutating more often than other regions.

Fingar said the bootcamp helped her continue to broaden computational skills for research. Ill be the first to admit Im not the best with computers, Fingar said. This opportunity has shown me that I am capable of learning and effectively utilizing them in a meaningful way that is relevant to my research. My next challenge will be to tackle the statistical programming language R.

Senior biochemistry major Charlotte Labrie-Cleary found the opportunity to work in remote teams and gain experience relevant to research were key takeaways.

I learned how to use incredibly powerful bioinformatic tools that I hope to learn more about in the future, Labrie-Cleary said. I learned about the evolution of viruses with a focus on coronaviruses. We learned in depth about the SARS-CoV-2 main protease as well as its spike protein and why theyre important. We learned about testing techniques for COVID-19 and how they work.

For Labrie-Cleary. learning so much at a fast pace was exhilarating and I feel lucky to have been able to participate, she said. It has shed light into the world of bioinformatics, which is something I have always been super interested in. This experience will give me a head start when considering graduate programs, and it excites me to learn more about it. As an undergraduate, I am fortunate to have been offered such a valuable experience, as many students at our level are not offered such during undergraduate studies.

Senior biochemistry major Michael Kirsch appreciated learning about topics such as the evolution of RNA viruses, development of testing for COVID-19, what parts of COVID-19 might be the best to target with medicines that are being developed to treat it, and how phylogenetic trees can be used to help piece together when different mutations in a virus branch off from one another, he said.

His team used Mol and Foldit to examine the protein 6YB7, the COVID-19 main protease, which could lead to research on what affects its ability to do its job as a protease, and future research can then be done on how to disrupt this protein from doing its job, Kirsch said.

Virology, the study of viruses, is among the future fields Kirsch is considering, and the bootcamp has further encouraged him. I now know how to use several new programs to visualize or otherwise analyze proteins, which will be useful in my last semester at Oswego, as the research I do with Dr. Koeppe is focused on determining the function of protein 3DL1, he said. Being able to better visualize it can only help my research efforts, which Im excited about.

The bootcamp will allow Koeppe to provide better lab experiences and topical opportunities for her students.

I am currently modifying some of the bootcamp materials to use them as online lab experiments in our biochemistry lab courses for the fall semester, and I will also create a unit on the novel coronavirus and the main protease for my masters-level enzymes course for the fall semester, Koeppe noted. Students who are interested in further study will be welcome to join my research group where they can begin with computational experiments to study the viral proteins with a goal of identifying a possible drug target.

Koeppe and chemistry faculty member Kestutis Bendinskas have been using tools developed at the boot camp to design experiments for studying SARS-CoV-2 in their biochemistry lab courses.

The experience can help Koeppe develop a unit on computational software for protein folding into our biochemistry lab curriculum that focuses on enzymes of unknown function, she said. The software we used for protein folding in the bootcamp was new to all of us, and we think it will be a good addition to what weve already been using in the lab.

As an Amazon Associate I earn from qualifying purchases.

See the original post:
Bootcamp allows Oswego researchers to explore fighting COVID-19 - NNY360

How a Google Engineer Used Her AI Smarts to Create the Ultimate Family Archive – PCMag UK

(Image: Getty)

COVID-19 lockdowns perhaps gave a few of you some time to organize old photos that have been languishing on SD cards or in boxes, but how many of you built an AI-powered searchable archive of family videos from almost 500 hours of footage?

Dale Markowitz, an Applied AI Engineer and Developer Advocate at Google, did just that. The Texas-based Princeton grad took hours of disorganized, miniDV tape footage housed on Google Drive and turned it into an archive "that let me search my family videos by memories, not timestamps," she wrote in a July blog post. It was the ultimate Father's Day gift.

We spoke with Markowitz recently to find out how machine learning helped her get it done, but why AI is only one part of the puzzle when it comes to solving complex problems.

Although this project used a raft of Google tools, which well get to, it was actually not for the day job, but the coolest Father's Day gift, right?[DM] At Google, I spend lots of time trying to think up new use cases for AI and build prototypes focused on the more business-y side. But I always wanted to work on more fun, zany stuff and, with quarantine, I finally had SO MUCH TIME. So, yes, this one was a gift for my dadwho, by the way, is also a huge programmer nerd who works in machine learning.

As your dad works in machine learning, he would totally get what it took to build it out. Let's go "under the hood" with the details.[DM] Sure. So I uploaded all of my dad's videos to a cloud storage bucket and then analyzed them with the Video Intelligence API, which returns JSON. Basically, the API does all the heavy lifting including: detecting scene changes; extracting text and timestamps on screen using computer vision; transcribing audio; tagging objects and scenes in images; and so on.

Because you needed to apply intelligence to what was probably hours of untagged material, right?[DM] Exactly. When my dad recorded on miniDV, the clips werent saved into separate files. They'd all be smashed into one long, three-hour recording, separated by little flashes of black and white. The API was able to pick out where those clip boundaries should have been.

Regarding audio transcription, that must have helped in tagging, categorizing, and identifying what was on all those miniDVs.[DM] Yes, and I found this to be the coolest part of the project, because it let me search for hyper specific things like "Pokemon" or "Gameboy." Also, my dad was a big video narrator, so I could search his commentary for milestones.

As an applied AI engineer, you're experienced in this field, but others using the API won't need to be up on machine learning, right? Essentially, it's not quite, but almost, out-of-the-box in terms of building out the metadata and intelligence?[DM] Confirmed. You dont need any ML expertise to build out this project. Its very developer-friendly. Having said that, there was one more AI part of this project, which was implementing search. I wanted to be able to search through all those transcripts, scene labels, and objects labels, but I didnt want to have to exactly match the words.

Because you needed a proper semantic search layer for this project?[DM] Exactly. I wanted to allow for near-matches and misspellings and even matching synonyms, such as treating the word trash the same as garbage." As you know, in semantic search," you want an algorithm that understands the semantic meaning of what youre saying regardless of the specific words and spellings you use. For that, I used a great Search as a Service tool called Algolia. I uploaded all my records (as JSON) and Algolia provided me with a smart semantic search endpoint to query those records.

Obviously, youve got a corporate account as a Googler to use all these tools. But what would the cost be for a non-Googler to do this? And are you sharing your GitHub codeyour GitHub code so people can replicate this?[DM] Yep, the code is all open source. Though I should add that a lot of these features are available through Google Photos, which works with videos too, apart from the ability to search transcripts. Cost-wise, I analyzed 126GB of video (about 36 hours) and my total cost was $300. I know that seems high, but it turns out the bulk of the cost came from one single type of analysisdetecting on-screen text. Everything else amounted to just $80. As on-screen text was the least interesting attribute I extracted, I recommend leaving that out unless you really need it. Also, the first 1,000 minutes of video falls in the Google Cloud free tier. Besides the ML parts, storing my data in Algolia runs me around $50 a month for around 90,000 JSON objects. But I havent done much optimizing, and they do have a free tier.

Youre the overall host on YouTube for the new series Making with Machine LearningMaking with Machine Learning." Whats up next there in terms of projects?[DM] Machine-generated recipes, automatically dubbed videos, and an AI dash cam. Well, if I can get those things to workI never really know until I start building them. Another thing Ive been fascinated with lately are ways to do machine learning with little or no data, and zero-shot learning. More on that coming soon.

Well look out for those. Now lets do some background on you: What drew you to study computer science and why specifically at Princeton?[DM] I originally decided to go to Princeton because I wanted to be a theoretical physicist, and I really admired Professor Richard Feynman when I was in high school.But back in 2013, when I was a sophomore in college, it really felt like computer science was the place to be: everything was developing so quicklyArduino, AI, brain-computer interfaces. In retrospect, though I didnt know it then, majoring in computer science was a great decision, because theres almost no field, scientific or otherwise, that hasnt benefited from machine learning. In fact, sometimes it seems like some of the most cutting-edge work in biology and neuroscience and physics is coming from ML.

Whats caught your eye recently in terms of ML?[DM] Specifically within the field of biophysics, Id say DeepMinds new protein folding model, AlphaFold 2, is a great example of ML.

You worked as a researcher on brain-machine interfaces to measure sustained attention. Can you give us a brief explanation of what you were doing there?[DM] In that lab, some researchers had discovered they could (roughly) measure attention by having people do an extremely mundane task in an fMRI machine and then analyzing their brain scans. They were actually using deep learning, which was pretty revolutionary in neuroscience at the time. The problem is that fMRI machines are extremely expensive. I was investigating whether you could get similar results using an EEG machine (which is much cheaper), and specifically a portable, wireless EEG (which is much much cheaper). The results were mixed, but I think, since then, portable EEG machines have gotten better at taking clear readings, and I have gotten better at machine learning.

You moved from data science to applied AI and your focus is on how people can apply AI, ML, etc. But do you also interface with the more theoretical AI people at Google too or only tangentially?[DM] There is a pretty tight relationship between Google Cloud and Google Research. The field changes so quickly that there has to be. When a splashy research paper comes out, it takes almost no time before customers start asking how to get it on Google Cloud. One good example is around explainability and responsible AI. Now that machine learning is becoming more accessible, more folks can build their own models. But how do you know those models are accurate? How do you know you can trust them, and that they wont make predictions that are embarrassing or offensive? The answer is closely linked to explainability, our ability to understand why models make the predictions they doi.e. its hard to trust black box models.

Yeah, theres a big push for explainable AIexplainable AI right now.[DM] This is a tough problem, and an active area of research across Google. But weve been working very closely with Google Research to add explainability into our customer-facing products.

At Google I/OGoogle I/O 2019, you focused on democratizing AIallowing developers to use Googles AI tools, like AutoML, and off-the-shelf APIs to create cool stuff. Tell us more about that. [DM] ML has gotten way easier and more accessible for developers over the past five years. And one of the reasons thats so exciting is because more people from different backgrounds start using it and we end up with very creative projects. Sometimes people see a project Ive built and theyll riff on it, which I think is super cool. For example, I built a tennis serve analyzer, and then some folks built a cricket and a badminton version. I saw a yoga pose detector, and someone built an AI Diary using some of the same tech as my video archive analyzer.

Thinking more broadly, it occured to me that many of your AI-powered projects are applications that could help non-neurotypical people to navigate the world. For example, you engineeredengineered an AI Stylist which could illuminate social cues and help people be workplace appropriate or situation appropriate.[DM] Interesting. On one hand, there are definitely great applications of AI for non-neurotypical folks. The most compelling one Ive heard of involves using computer vision to understand facial expressions and emotions. On the flip side, I try to avoid using machine learning in situations where the result of a mistake is catastrophic.

On that note, when I interviewed Dr. Janelle Shaneinterviewed Dr. Janelle Shane, she had some bizarre brownie recipes generated by one of her AIs, because that stuff is harder than most people imagine. For example, AI doesnt have common sense," so you had to build in rules that a human wouldnt need - i.e. I need two shoes, a left one and a right one, but only one shirt or hat." Any wardrobe mishaps with the stylist before it got it right?[DM] Oh yes, 100%. Furthermore, I would say using a combination of ML and human rules is a pretty good design pattern. One mistake I see people make a lot is try to completely, end-to-end solve a problem with AI. Its better to use ML only for the parts of your system that really need it, such as recognizing a clothing item from an image. But then writing simple rules in places where ML isnt necessarysuch as combining clothing items to make an outfit. Human rulesi.e. An outfit contains exactly two shoesare usually easier to understand, debug, and maintain than ML models. One thing that seemed to trip up the stylist app was that I took a bunch of pictures of clothing on mannequins; my vision model was trained on pictures of people, not mannequins.

The vision model which was looking for humans not static clothes horses?[DM] Yup. That really tricked the model. It was convinced the mannequin was a suitcase or something. By the way, I published the code on GitHub if others want to try it out.

At Google I/O, you also talked about the custom sentiment analysis using natural language. Has that been deployed into something cool like a concurrent translator that can detect irony or emotioni.e. good for non-native speakers while on business trips abroadif we ever get to do those again?[DM] Interesting idea. Were still struggling with irony detection in NLP. But can you really blame a computer for not recognizing irony when lots of humans cant, either?

Good point.[DM] I also suspect irony is largely contextuali.e. text paired with an image, or spoken in a particular way, which makes the problem more challenging. Detecting emotion from speech is a cool idea. But Id probably opt not to analyze just the words the person is saying (text sentiment) and focus more on their intonation. Sounds like a neat project. But like many ML problems, the challenge is finding a good training dataset.

True. So, wrapping it up, do you see the AI tools that youre working with now are a way of building a smart layer between IRL and our silicon cousins (embodied/non-bodied AIs)? For example, when I interviewed AI researcher Dr. Justin Liinterviewed AI researcher Dr. Justin Li, we talked about AI being able to anticipate our needs before we know we have them. [DM] In the future, yes, I think humans and AIs will work closely together. But for me whats most compelling are use cases where machine learning models are uniquely well-suited to do something that humans cant do or arent good at. For example, people make really good assistants and companions and teachers, but theyre not very good at processing millions of web pages in seconds or discovering exoplanets or predicting how proteins fold. So its in these applications, I believe, that AI can make the most impact.

Read more from the original source:
How a Google Engineer Used Her AI Smarts to Create the Ultimate Family Archive - PCMag UK

GT Gain Therapeutics SA Announces Funding from the Swiss Innovation Agency Supporting a 3-year Research Collaboration Project with the Institute for…

- Researchers will further develop the Site-directed Enzyme Enhancement Therapy (SEE-Tx) technology for the treatment of rare genetic and neurodegenerative diseases

- The collaborative agreement unites resources from the Institute for Research in Biomedicine (IRB)-USI; Neurocentro -Ente Ospedaliero Cantonale (EOC) & GT GAIN Therapeutics, SA

LUGANO, Switzerland, Dec. 15, 2020 (GLOBE NEWSWIRE) -- GT Gain Therapeutics SA (Gain), a subsidiary of Gain Therapeutics, Inc.,a biotechnology company focused on redefining drug discovery by identifying and optimizing allosteric binding sites that have never before been targeted, along with the Institute for Research in Biomedicine (IRB, affiliated to USI Universit della Svizzera Italiana) and the Neurocentro announced today that Innosuisse, the Swiss Innovation Agency, has agreed to support the CHF 1.5M project by funding approximately CHF 850,000 to leverage these world class research organizations and promote continued innovation in the area of CNS diseases. The remaining support will come from Gain to cover the cost of related headcount expenses being dedicated to the project. The award specifically supports further investigation of the mechanisms of action of Gains proprietary STAR small molecule therapeutic candidates on lysosomal dysfunction and prion-like transmission of toxic forms of protein aggregates associated with neurodegenerative diseases.

Being recognized as an Innosuisse funded innovation project reinforces the support for our innovative approach and unites us with scientists and researchers as passionate as we are to discover new therapeutic approaches using our SEE-Tx target identification platform, said Manolo Bellotto, Ph.D., President and General Manager of Gain. The specific know-how in protein quality control by Prof. Molinari at the IRB and the expertise in neurosciences of Dr. Paganetti from Neurocentro will certainly contribute to a further understanding of the mechanism of action of our molecules in rare and genetic diseases, thus accelerating their development towards the clinic.

Dr.Maurizio Molinari, group leader of the Protein Folding and Quality Control research team from the IRB added, We are honored to be collaborating with the Gain team and to evaluate Gains novel therapeutic candidates as we work to advance new, innovative treatment options for rare lysosomal disorders and neurodegenerative diseases for which there are currently few treatment options. We are grateful to the Swiss Innovation Agency for their support and look forward to initiating this critical research program.

About Gain Therapeutics, Inc.

Gain Therapeutics, Inc. is redefining drug discovery with its SEE-Tx target identification platform. By identifying and optimizing allosteric binding sites that have never before been targeted, Gain is unlocking new treatment options for difficult-to-treat disorders characterized by protein misfolding. Gain was originally established in 2017 with the support of its founders and institutional investors such as TiVenture, 3B Future Health Fund (formerly known as Helsinn Investment Fund) and VitaTech. It has been awarded funding support from The Michael J. Fox Foundation for Parkinsons Research (MJFF) and The Silverstein Foundation for Parkinsons with GBA, as well as from the Eurostars-2 joint program with co-funding from the European Union Horizon 2020 research and Innosuisse. In July 2020, Gain Therapeutics, Inc. completed a share exchange with GT Gain Therapeutics SA., a Swiss corporation, whereby GT Gain Therapeutics SA became a wholly owned subsidiary of Gain Therapeutics, Inc. For more information, visit https://www.gaintherapeutics.com/

About the Institute for Research in Biomedicine (IRB)

The Institute for Research in Biomedicine was founded in 2000 with the clear and ambitious goal of advancing the study of human immunology, with particular emphasis on the mechanisms of host defense. The activities of the 13 research groups now extend beyond immunology to include the fields of DNA repair, rare diseases, structural and cell biology. Located in Bellinzona, capital of the Italian-speaking Canton of Ticino, the IRB is an affiliated institute of the USI Faculty of Biomedical Sciences. For more information, visit : http://www.irb.usi.ch

About Neurocentro -Ente Ospedaliero Cantonale (EOC)

The EOC multisite hospital is organized and managed as a modern company at the service of the patient. It has structures with clear segregations of functions and flexible management systems that foster innovation, accountability and simplification.Our approach favors a collegial and participatory management style. General management and hospital directors form the EOC Management Coordination Conference, physicians are directly involved in EOC management through the Clinical Coordination Conference. The other professional categories actively participate in the management of the EOC within inter-hospital groups.For more information, visit http://www.eoc.ch/en/Centri-specialistici/NSI/NSI.html

Forward-Looking Statements

Any statements in this release that are not historical facts may be considered to be forward-looking statements. Forward-looking statements are based on managements current expectations and are subject to risks and uncertainties which may cause results to differ materially and adversely from the statements contained herein. Such statements include, but are not limited to, statements regarding Gain Therapeutics, Inc. (Gain) expected use of the proceeds from the Series B financing round; the market opportunity for Gains product candidates; and the business strategies and development plans of Gain. Some of the potential risks and uncertainties that could cause actual results to differ from those predicted include Gains ability to: make commercially available its products and technologies in a timely manner or at all; enter into other strategic alliances, including arrangements for the development and distribution of its products; obtain intellectual property protection for its assets; accurately estimate its expenses and cash burn and raise additional funds when necessary. Undue reliance should not be placed on forward-looking statements, which speak only as of the date they are made. Except as required by law, Gain does not undertake any obligation to update any forward-looking statements to reflect new information, events or circumstances after the date they are made, or to reflect the occurrence of unanticipated events.

Gain Therapeutics Investor Contact:Daniel FerryLifeSci Advisors+1 617-430-7576daniel@lifesciadvisors.com

Gain Therapeutics Media Contact:Cait Williamson, Ph.D.LifeSci Communications+1 646-751-4366cait@lifescicomms.com

Read more here:
GT Gain Therapeutics SA Announces Funding from the Swiss Innovation Agency Supporting a 3-year Research Collaboration Project with the Institute for...

Those We Lost in 2020 – The Scientist

For a complete list of our obituaries, seehere.

Jeff McKnight, a molecular biologist at the University of Oregon, died in October at the age of 36.

McKnights research focused on chromatin, a complex of DNA and proteins that controls when and how DNA can be accessed for replication and gene expression. He was one of the earliest researchers in the world capable of directly manipulating its structure, stemming back to his postdoctoral work at the Fred Hutchinson Cancer Research Center using the model organism Saccharomyces cerevisiae. When he had started his own lab in 2016, McKnight said at the time that his real dream was to apply his work to the dozens of human diseases that involve some level of chromatin disruption, including Parkinsons, Alzheimers, and Huntingtons.

Prior to his death from lymphoma, McKnight had spent months chronicling his diagnosis and treatment on social media, prompting an outpouring of support from fellow scientists. He had this humility and vulnerability about him that was really endearing, David Garcia, a molecular biologist at the University of Oregon, told The Scientist.

Biologist Lynika Strozier, a researcher at the Field Museum and an instructor at Malcom X College, died June 7 at age 35 due to complications associated with COVID-19.

After being introduced to molecular biology as an undergraduate at Truman College, Strozier developed a passion for using DNA to identify new and sometimes cryptic species. For her thesis work as a masters student at Loyola University, Strozier sequenced DNA from 200 individual birds in Madagascar thought to belong to three species. Instead, she identified several new species that were indistinguishable based only on the birds appearance.

Her steady hand and aptitude in extracting usable genetic material from old samples earned the admiration of her colleagues. Our entire team entrusted Lynika with extracting DNA from old dried plant material of over 15 years and only very little material from which to do so, Matt Von Konrat, the head of botanical collections at the Field Museum, told The Scientist.

VANDERBILT UNIVERSITY MEDICAL CENTER

Nobel laureate and biochemist Stanley Cohen, who led pioneering studies of cell growth factors, died in February. He was 97.

Stans work not only provided key insights into how cells grow, but it led to the development of many drugs that are used to treat cancer, Lawrence Marnett, the dean of basic sciences at Vanderbilt University, where Cohen taught for 40 years, said to The Tennessean.

Cohens work on different types of growth factors alongside biochemist Rita Levi-Montalcini earned them the 1986 Nobel Prize in Physiology or Medicine. Cohen was honored for his discovery of epidermal growth factora protein that stimulates cell growth and differentiation and plays an important role in tumor developmentwhile Levi-Montalcini was acknowledged for first isolating nerve growth factor. Growth factor receptors have since become the target of numerous drugs, such as gefitinib and cetuximab, that slow or prevent the progression of certain cancers.

SAMARA VISE, KOCH INSTITUTE AT MIT

Angelika Amon, a cell biologist at MIT, died on October 29 from ovarian cancer at the age of 53.

Amon dedicated her career to researching the cell cycle and how disruptions to its normal function can lead to cancer.

During her PhD at the University of Vienna and her subsequent postdoc at MITs Koch Institute for Integrative Cancer Research, Amon used model organisms such as yeast and fruit flies to study how certain proteins and enzymes direct cells through mitosis.

Later, Amon turned her focus to the study of aneuploidy, an abnormal number of chromosomes, and chromosome segregation. She found that extra chromosomes disrupt protein folding and cell metabolism, leading to errors in those processes that can drive cancer.

More than anyone else Ive ever met, she was an absolute force of nature, Matthew Vander Heiden, an MIT biologist and close friend of Amon, told The Scientist. She just has this larger than life personalitytheres no other way to put it.

WILL KIRK/JOHNS HOPKINS UNIVERSITY

Computational biologist James Taylor, who developed a widely used bioinformatics platform, died in April. He was 40.

James made huge contributions to open-source, accessibility, and reproducibility, genomicist Andrew Carroll tweeted following his death. Anyone who runs a bioinformatics tool on the cloud does so thanks to Jamess work.

During his PhD at Penn State University, Taylor helped develop the Galaxy Project, a platform that allows researchers to share genomic data without needing to know how to program. He continued refining the platform as he moved from teaching at Emory University to Johns Hopkins University, and since then Galaxy has been used in more than 10,000 publications across disciplines. Prior to his death, Taylor spoke on Twitter of the need to make transparent, reusable and reproducible analysis pipelines to address the current pandemic, by developing resources for best practices in sharing and analyzing data.

ED SOUZA/STANFORD NEWS SERVICE

Sleep scientist William Dement, who described a number of sleep disorders and opened one of the worlds first sleep disorder clinics, died in June. He was 91.

During his graduate studies at the University of Chicago in the 1950s, Dement studied the physiology of REM sleep and its relationship to dreaming. He later joined the faculty at Stanford University, where he taught for 50 years. There, his focus became the study of sleep apnea and the effects of sleep deprivation. In 1970, he launched the Stanford Sleep Medicine Center and is credited with prompting Congress to establish the National Center on Sleep Disorders Research.

There are not a lot of people who can say they saved the lives of hundreds of thousands of people, Emmanuel Mignot, a professor of psychiatry and behavioral sciences at Stanford University, said in an obituary. But just by pushing this field forward, making sleep apnea recognized, as well as sleep disorders and sleep deprivation, Bill did that.

Wendy Havran, an immunologist at the Scripps Research Institute who studied the role of gamma-delta T cells in wound healing, died January 20 at the age of 64.

Havran first became interested in immunology after meeting John Cambier, an immunologist at Duke University, where she completed her undergraduate degree. While she had intended to study medicine, she became enamored of doing research. It just clicked, and there was no going back, she told The Scientist in a 2019 profile. I wanted to understand how the immune system worked.

During her doctorate research at the University of Chicago, Havran used monoclonal antibodies to study CD4 and CD8 surface markers on T cells. Later, as a postdoc at the University of California, Berkeley, Havran began focusing specifically on gamma-delta T cells, which had only just been described. She was able to map their abundance throughout the body, showing for the first time that they were common in the skin and intestines. In her own lab at Scripps, Havran went on to demonstrate the cells ability to heal wounds and suppress tumor growth.

JOHNS HOPKINS BLOOMBERG SCHOOL OF PUBLIC HEALTH

Immunologist and microbiologist Noel Rose, whose early experiments established the concept of autoimmunity, died of a stroke on July 30 at the age of 92.

Before his pioneering work, it was believed that the body was incapable of launching an immune response against itself. But as a young medical student at the University of Buffalo, Rose showed that rabbits injected with their own thyroid-derived antigens mounted an immune response that damaged or destroyed the animals thyroid. Over the next several decades, he would further unravel the causes and mechanisms of autoimmune diseases, publishing more than 880 articles and book chapters.

In every aspect, [Rose] is the father of autoimmunity, George Tsokos, a professor of rheumatology at Harvard Medical School, told The Scientist in a profile of Rose. The man opened a whole chapter in the book of medicine.

LIZA GREEN, HARVARD MEDICAL SCHOOL

Phillip Leder, a molecular geneticist at Harvard Medical School whose research furthered the fields of molecular biology, immunology, and cancer genetics, died in February. He was 85.

Working alongside NIH geneticist Marshall Nirenberg as a postdoc in the 1960s, Leder developed a technique that confirmed, for the first time, that amino acids were encoded by three nucleotides. Speaking in a 2012 interview, he recalled the excitement of those early experiments. I would go to bed thinking about the next days experiments and then jump out of bed in the morning and rush to the laboratory. It was a lot of work, but the intellectual excitement was enormous.

Having revealed the genetic basis of protein coding, Leder next went on to map the first complete sequence of a mammalian gene, develop the first recombinant DNA vector system, and discover a cancer-causing gene that led to the development of the first mouse model of cancer, among other achievements. He established the genetics department at Harvard Medical School in 1981 only a year after joining the faculty and served as chair for 25 years.

Phil Leder was special. Among great scientists, he was special, and among scientists, he was an icon, David Livingston, a geneticist at Harvard, told The Scientist.

University of california, san diego

Molecular virologist Flossie Wong-Staal, a researcher at the University of California, San Diego (UCSD), who first cloned the human immunodeficiency virus (HIV), died in July at age 73 due to complications from pneumonia.

When Wong-Staal first entered the laboratory of fellow virologist Robert Gallo as a postdoc in 1973, scientists were skeptical that retroviruses could cause cancer in humans. Wong-Staals work helped to overturn this dogma after she and her team identified the first human retrovirus (HTLV-1) and showed that it could indeed lead to cancer. Together, she and Gallo published more than 100 papers in 20 years, making Wong-Staal the most-cited woman in science during the 1980s.

As AIDS cases began to spike in the 1980s, Wong-Staal became the first person to clone HIV the retrovirus that causes the diseaseand began studying the functions of its genes, a necessary step towards developing eventual treatments. She left Gallos lab at the National Cancer Institute in 1990 to launch the Center for AIDS Research at UCSD, where she spent the next several decades studying the virus and developing treatments, many of which are still in use today.

Go here to read the rest:
Those We Lost in 2020 - The Scientist

Has Google’s DeepMind revolutionized biology? | TheHill – The Hill

Every budding biologist learns about proteins and the amino acids that build them. Proteins are the building blocks of life, but knowing the sequence for the protein is only half of the story. How the protein folds onto itself determines what sections are exposed and can interact with other molecules, and therefore also what sections are hidden.

This is called the protein folding problemand has stumped the scientific community for about 50 years. Scores of researchers around the world are working to predict how proteins are folded, many using artificial intelligence (AI).

Biologists want to be able to predict how a protein folds because that gives insight into what it does and how it functions in the body. Geneticists and researchers have gained understanding about genes that encode for proteins, but experts have less knowledge about what happens when proteins are released to do their jobs.

One group at DeepMind, a Google AI offshoot, built an AI system that has done what others have not been able to. The group entered their algorithm, called AlphaFold, in the biennial protein-structure prediction challenge called Critical Assessment of Structure Prediction (CASP). The organizers of CASP look at the accuracy of predictions to assess how good the solutions are. The assessment is done blind, meaning the assessors dont know whose results they are looking at.

BREAKING NEWS ON THE CORONAVIRUS PANDEMIC

CDC CUTS LENGTH OF COVID-19 QUARANTINE TIME AFTER EXPOSURE

UK BECOMES FIRST WESTERN NATION TO AUTHORIZE COVID-19 VACCINE

CDC DECIDES WHO WILL RECEIVE FIRST DOSES OF COVID-19 VACCINES

CORONAVIRUS EPIDEMIC WAS SPREADING IN US LAST CHRISTMAS, LONG BEFORE IT WAS IDENTIFIED IN CHINA, NEW STUDY FINDS

This year, AlphaFold has come out on top, beating its past performance and others in the competition.

This is a big deal,said John Moult, who is a computational biologist at the University of Maryland in College Park and co-founded CASP in 1994, to Nature. In some sense the problem is solved.

America is changing faster than ever! Add Changing America to your Facebook or Twitter feed to stay on top of the news.

Research groups that dont use AI usually focus on experiments and collect data like X-ray diffraction data. One group that was trying to figure out a bacteria protein has been studying it for a decade while AlphaFold solved it in half an hour, according to Nature.

This is a problem that I was beginning to think would not get solved in my lifetime,said Janet Thornton, who is a structural biologist at the European Molecular Biology Laboratory-European Bioinformatics Institute and a past assessor for CASP, to Nature.

DeepMind is mostly known for its success in chess, Go and other games. Demis Hassabis, DeepMinds founder and chief executive,said to The Guardian, These algorithms are now becoming mature enough and powerful enough to be applicable to really challenging scientific problems.

READ MORE LIKE THIS FROM CHANGING AMERICA

LIQUID BIOPSIES COULD LEAD TO EARLY CANCER DETECTION

CORONAVIRUS EPIDEMIC WAS SPREADING IN US LAST CHRISTMAS, LONG BEFORE IT WAS IDENTIFIED IN CHINA, NEW STUDY FINDS

SEVERAL DIFFERENT TYPES OF DEPRESSION ARE SET TO COLLIDE THIS WINTER

WOMEN OF COLOR ARE TIPPING THE BALANCE OF POWER IN U.S. CITIES

Read more:
Has Google's DeepMind revolutionized biology? | TheHill - The Hill

Deep medicine: Artificial intelligence is changing the face of healthcare, daily – Yiba

Professor Tshilidzi Marwala is the Vice-Chancellor and Principal of the University of Johannesburg. He recently penned an opinion article that first appeared in theDaily Maverickon 07 December 2020.

This year has been a great definer. As we waged a battle against an unknown entity, proponents of artificial intelligence (AI) were swift to act. Just last week, DeepMind announced that it has cracked what is referred to as a 50-year-old scientific riddle. It has solved the protein-folding problem. In other words, it can determine a proteins 3D shape from its amino-acid sequence, making it easier to develop treatments for a range of diseases from cancer to the coronavirus.

To do this, researchers trained the DeepMind algorithm on a public database, which contained about 170,000 protein sequences and their shapes over a few weeks, running the equivalent of 100 to 200 graphics processing units. In recent years, DeepMind has been most recognised for its ability to beat human beings in games such as Go or Atari Classics. These were, in a sense, testing grounds for ultimately solving real-world problems.

As DeepMinds founder Demis Hassabis said at the announcement last week: It marks an exciting moment for the field. These algorithms are now becoming mature enough and powerful enough to be applicable to really challenging scientific problems. In fact, many had expected this kind of advancement in AI only in a few decades from now.

This indicates the advent of the Fourth Industrial Revolution (4IR) the era we find ourselves in, where intelligent technologies permeate all aspects of our lives. AI, which is the most significant technology of the 4IR, is already changing how we live, work and communicate by reshaping government, education, healthcare and commerce. In his bookDeep Medicine,Eric Topol distinguishes between shallow and deep medicine. Shallow medicine is a healthcare system based on observations of community groups (for example, people of African descent have a higher risk of prostate cancer than other community groups), whereas deep medicine is based on individualised medicine that is enabled by AI.

Not only do we have more access to information than ever before, but we also see a confluence of cyber, physical and biological technologies that no longer exist in labs, but impact on us every day. Proponents have long argued that the 4IR could be the key to finding solutions to some of our most deep-seated problems. The unprecedented responses to the coronavirus pandemic have been an exemplification of this.

For instance, AI has made the detection of the coronavirus easier. Alibabas research institute, Damo Academy, has developed an AI algorithm that can detect the coronavirus in just under 20 seconds with 96% accuracy. The AI was trained using 5,000 samples from confirmed cases and can detect the virus from chest CT scans, differentiating between infected patients and general viral pneumonia cases.

South Korea was swift to act following the outbreak in China, anticipating a spread into its borders. The government organised the private sector to develop testing kits for the virus. Molecular biotech company Seegene in South Korea used AI to accelerate these kits development. This facilitated the submission of its solution to the Korea Centers for Disease Control and Prevention (KCDC) only three weeks after scientists began working on this solution. Under normal situations, this process would have taken two to three months with an approval process of about 18 months.

It is not just pockets of AI that have cropped up in these regions. The opportunity for AI to speed up the implementation of vaccines, drugs and diagnostics is gaining traction elsewhere. Projects such as the Covid-19 Open Research Dataset provide free access to the texts of almost 25,000 research papers, while the Covid-net open access neural network is working on systems similar to those deployed by the Damo Academy.

Companies such as BenevolentAI, based in the United Kingdom, are using AI and the available data to scour through existing drugs that could be used to treat coronavirus patients until a vaccine becomes available.

Vir Biotechnology and Atomwise, start-ups in the United States, are using algorithms to identify a molecule that could facilitate treatment. Now, as various vaccines are in the final testing stages, algorithms are being used to sift through data on potential adverse reactions. Companies such as Genpact UK have signed contracts with the UK government to ensure that nothing is missed as preparations begin for mass vaccinations in the coming year. This is significant given the rapid timeline in which many of these vaccines have been developed and the various unknowns that remain.

AI solutions once thought of as futuristic and unrealistic are now commonplace. We see far more advances than we had expected at this stage, perhaps indicating the urgency that the pandemic has presented.

Similarly, there has been a shift to find AI solutions in Africa. Data science competition platform Zindi which is based in South Africa and Ghana has initiated a competition sponsored by the Artificial Intelligence for Development-Africa Network (AI4D-Africa), which requires data scientists to create an epidemiological model that forecasts the spread of Covid-19 throughout the globe. This is critical for both policy makers and health workers to make informed decisions and take action.

In Kenya, start-up Afya Rekod deploys AI and Blockchain to establish a health-data platform that lets users store their health records, access health information and connect to health service providers.

Of course, it is not only in the context of the coronavirus pandemic that there have been AI advances. There have been great strides in bridging some of the inequalities that exist in the healthcare system. In Rwanda, for instance, the government has collaborated with US start-up Zipline to deliver blood supplies by drones to remote areas. Where a journey would have taken three hours by car, a drone can complete the trip within six minutes. This addresses emergency medical supply requirements in rural areas.

Just last month, to improve access and quality of services to rural communities in South Africa, the Department of Health in Limpopo installed CT-Scans and Picture Archiving Communication System (PACS) in the province. The availability of this equipment at regional hospitals now improves the speed of diagnosis and management of the associated conditions and indicates an embracing of the 4IR.

This is vital because according to the General Household Survey conducted by Statistics SA, only 17% of South Africans have medical insurance, the critical key for private healthcare. About 82% of South Africans fall outside the medical-aid net, and, as a result, are largely dependent on public healthcare. According to Statistics South Africa, in 2017, 81% of households that used public healthcare services were satisfied or very satisfied with public facilities services.

AI also addresses concerns of a shortage of doctors, particularly in the public sector. For example, the increased speed and accuracy of cancer diagnostics through analytics which can characterise tumours and predict therapies has not replaced doctors, but quickened their efforts and given them the space to attend to more patients. Technologies such as AI will decrease the cost of health care globally.

Almost two-thirds of healthcare costs are from non-communicable diseases such as cancer, strokes, heart failure and kidney failure that can be treated more effectively and at less cost if diagnosed early.

For example, in China, a company called Infervision developed AI algorithms that efficiently and accurately read medical images to augment radiologists in diagnosing cancer.

As Dhruv Khullar, a physician at New York-Presbyterian Hospital, said, most fundamentally, it means recognising that humans, not machines, are still responsible for caring for patients. It is our duty to ensure that we are using AI as another tool at our disposal not the other way around.

AI solutions once thought of as futuristic and unrealistic are now commonplace. We see far more advances than we had expected at this stage, perhaps indicating the urgency that the pandemic has presented.

What is clear is that, like many other sectors, health care will be transformed by AI and we need to ready ourselves for these shifts. As Enrico Coiera aptly put it inThe Lancetin 2018, what is the fate of medicine in the time of AI? Our fate is to change.

*The views expressed in the article is that of the author/s and does not necessarily reflect that of the University of Johannesburg.

Source: UJ

More:
Deep medicine: Artificial intelligence is changing the face of healthcare, daily - Yiba

Tech.eu Podcast #198: Even more money for e-scooters, new VC funds, protein folding, and we talk to Sebastian Peck of InMotion Ventures – Tech.eu

The Tech.eu Podcast is a show in which we discuss some of the most interesting stories from the European technology scene and interview leading entrepreneurs and investors from across the region.

This week, we talk about whats going on in European tech, including some of the biggest funding rounds of the week, new VC funds, science and research news, and much more. Weve also spoken to Sebastian Peck, managing director of InMotion Ventures.

You can find the latest episode embedded below. Subscribe today and dont miss new episodes:

And here are the notes and links for this weeks episode:

Voi, the European micromobility rental company, raises $160 million additional equity and debt funding

UK-based HungryPanda raises $70 million to expand its online Asian food delivery business worldwide

Monzo, the UK challenger bank, picks up additional 60 million in funding

UK edtech startup MEL Science snags $14 million Series B

SoftBank buys 10.1 percent stake in Sinch after its meteoric surge

This is where Target Global wants to invest its new 300+ million fund

Firstminute Capital launches second $111 million fund, featuring a whos-who of founders as LPs

The European Investment Bank Group debuts new 150 million financing instrument to support European AI tech firms

London AI lab claims breakthrough that could accelerate drug discovery

Interview with Sebastian Peck, managing director of InMotion Ventures, a firm backed by Jaguar Land Rover

We hope you enjoy(ed) the podcast! Please feel free to email us with any questions, suggestions, and opinions topodcast@tech.eu or tweet at us @tech_eu.

Image credit: National Cancer Institute on Unsplash

Link:
Tech.eu Podcast #198: Even more money for e-scooters, new VC funds, protein folding, and we talk to Sebastian Peck of InMotion Ventures - Tech.eu

Will AI empower scientists or replace them? – Techerati

Googles DeepMind AI team solved a long-running biological problem

Scientists are not about to lose their jobs to more sophisticated artificial intelligence instead it will help them work even better, an expert in the field has said following a Google breakthrough.

Last week, the tech giants DeepMind AI specialists based in the UK made a leap forward in solving one of biologys biggest challenges, the five-decade-old protein folding problem.

Determining the structure of a protein opens up a world of possibilities, from understanding neurological diseases like Parkinsons, to discovering new drugs.

The problem is there are so many and it takes time to understand them all we have only managed to unfold a fraction of the millions of known proteins in living things.

But what does this mean for scientists going forward?

Like many jobs touched by technology, it does not mean their skills will no longer be needed, according to Dr Aldo Faisal, professor of AI and neuroscience at Imperial College London.

Instead it will cut down on mundane tasks, allow research to be carried out faster, and enable scientists to concentrate on more in-depth experiments.

I think what were going to see is that AI is going to empower scientists, its not about replacing scientists, its about empowering them to be able to do more and effectively taking away the boring parts of the work so to speak that are routine and mundane and allowing them to move quicker, discover things faster and I think thats one of the biggest appeals of AI, Dr Faisal told the PA news agency.

The protein folding and AlphaFold is beautiful because it shows that one can test hypotheses much, much quicker than with current conventional technologies about how protein folds and of course how protein folds tell us something about how they can function, interact and so this will basically save time and allow people to very quickly explore protein structures without having to do costly and slow great experiments.

Although AI has been used to revolutionise science for several years, Dr Faisal said we are seeing loads of other applications arrive and he expects more to come.

For example, earlier this year a group of scientists from Massachusetts Institute of Technology (MIT) used AI to help them uncover new types of powerful antibiotics, capable of killing some of the worlds most problematic disease-causing bacteria.

That was a very fortuitous discovery they made using AI and were seeing loads of other applications in understanding, basically, bringing together data about health care and environment and the context in which people live in relating that to the genes and the function of proteins inside their body, Dr Faisal continued.

Establishing these links, basically connecting healthcare data, connecting daily life data,

See the article here:
Will AI empower scientists or replace them? - Techerati

Genesis Therapeutics raises $52M A round for its AI-focused drug discovery mission – TechCrunch

Sifting through the trillions of molecules out there that might have powerful medicinal effects is a daunting task, but the solution biotech has found is to work smarter, not harder. Genesis Therapeutics has a new simulation approach and cross-disciplinary team that has clearly made an impression: the company just raised a $52 million A round.

Genesis competed in the Startup Battlefield at Disrupt last year, impressing judges with its potential, and obviously others saw it as well in particular Rock Springs Capital, which led the round.

Over the last few years many companies have been formed in the drug discovery space, powered by increased computing and simulation power that lets them determine the potential of molecules in treating certain diseases. At least thats the theory. The reality is a bit messier, and while these companies can narrow the search, they cant just say here, a cure for Parkinsons.

Founder Evan Feinberg got into the field when an illness he inherited made traditional lab work, as an intern at a big pharma company, difficult for him. The computational side of the field, however, was more accessible and ended up absorbing him entirely.

He had dabbled in the area before and arrived at what he feels is a breakthrough in how molecules are represented digitally. Machine learning has, of course, accelerated work in many fields, biochemistry among them, but he felt that the potential of the technology had not been tapped.

I think initially the attempts were to kind of cut and paste deep learning techniques, and represent molecules a lot like images, and classify them like youd say, this is a cat picture or this is not a cat picture, he explained in an interview. We represent the molecules more naturally: as graphs. A set of nodes or vertices, those are atoms, and things that connect them, those are bonds. But were representing them not just as bond or no bond, but with multiple contact types between atoms, spatial distances, more complex features.

The resulting representation is richer and more complex, a more complete picture of a molecule than youd get from its chemical formula or a stick diagram showing the different structures and bonds. Because in the world of biochemistry, nothing is as simple as a diagram. Every molecule exists as a complicated, shifting 3D shape or conformation where important aspects like the distance between two carbon formations or bonding sites is subject to many factors. Genesis attempts to model as many of those factors as it can.

Step one is the representation, he said, but the logical next step is, how does one leverage that representation to learn a function that takes an input and outputs a number, like binding affinity or solubility, or a vector that predicts multiple properties at once?

Thats the work theyve focused on as a company not just creating a better model molecule, but being able to put a theoretical molecule into simulation and say, it will do this, it wont do this, it has this quality but not that one.

Some of this work may be done in partnerships, such as the one Genesis has struck up with Genentech, but the teams could very well find drug candidates independent of those, and for that reason the company is also establishing an internal development process.

The $52 million infusion ought to do a lot to push that forward, Feinberg wrote in an email:

These funds allow us to execute on a number of critical objectives, most importantly further pioneering AI technologies for drug development and advancing our therapeutics pipeline. We will be hiring more top notch AI researchers, software engineers, medicinal chemists and biotech talent, as well as building our own research labs.

Other companies are doing simulations as well and barking up the same tree, but Feinberg says Genesis has at least two legs up on them, despite the competition raising hundreds of millions and existing for years.

Were the only company in the space thats working at the intersection of modern deep neural network approaches and biophysical simulation conformational change of ligands and proteins, he said. And were bringing this super technical platform to experts who have taken FDA-approved drugs to market. Weve seen tremendous value creation just from that the chemists inform the AI too.

The recent breakthrough of AlphaFold, which is performing the complex task of simulation protein folding far faster than any previous system, is as exciting to Feinberg as to everyone else in the field.

As scientists, we are incredibly excited by recent progress in protein structure prediction. It is an important basic science advance that will ultimately have important downstream benefits to the development of novel therapeutics, he wrote. Since our Dynamic PotentialNet technology is unique in how it leverages 3D structural information of proteins, computational protein folding similar to recent progress in cryo-EM is a nice complementary tailwind for the Genesis AI Platform. We applaud all efforts to make protein structure more accessible such that therapeutics can be more easily developed for patients of all conditions.

Also participating in the funding round were T. Rowe Price Associates, Andreessen Horowitz (who led the seed round), Menlo Ventures and Radical Ventures.

Read more from the original source:
Genesis Therapeutics raises $52M A round for its AI-focused drug discovery mission - TechCrunch